Kosteneffectiviteit van 10- en 13-valent geconjugeerde pneumokokkenvaccins bij kinderen. KCE reports 155A

Transcriptie

1 Kosteneffectiviteit van 10- en 13-valent geconjugeerde pneumokokkenvaccins bij kinderen KCE reports 155A Federaal Kenniscentrum voor de Gezondheidszorg Centre fédéral d expertise des soins de santé 2011

2 Het Federaal Kenniscentrum voor de Gezondheidszorg Voorstelling: Het Federaal Kenniscentrum voor de Gezondheidszorg is een parastatale, opgericht door de programma-wet van 24 december 2002 (artikelen 262 tot 266) die onder de bevoegdheid valt van de Minister van Volksgezondheid en Sociale Zaken. Het Centrum is belast met het realiseren van beleidsondersteunende studies binnen de sector van de gezondheidszorg en de ziekteverzekering. Raad van Bestuur Effectieve leden: Plaatsvervangers: Regeringscommissaris: Pierre Gillet (Voorzitter), Dirk Cuypers (Ondervoorzitter), Jo De Cock (Ondervoorzitter), Frank Van Massenhove (Ondervoorzitter), Maggie De Block, Jean-Pierre Baeyens, Ri de Ridder, Olivier De Stexhe, Johan Pauwels, Daniel Devos, Jean-Noël Godin, Xavier De Cuyper, Palstermans Paul, Xavier Brenez, Rita Thys, Marc Moens, Marco Schetgen, Patrick Verertbruggen, Michel Foulon, Myriam Hubinon, Michael Callens, Bernard Lange, Jean-Claude Praet. Rita Cuypers, Christiaan De Coster, Benoît Collin, Lambert Stamatakis, Karel Vermeyen, Katrien Kesteloot, Bart Ooghe, Frederic Lernoux, Anne Vanderstappen, Greet Musch, Geert Messiaen, Anne Remacle, Roland Lemeye, Annick Poncé, Pierre Smiets, Jan Bertels, Celien Van Moerkerke Yolande Husden, Ludo Meyers, Olivier Thonon, François Perl. Yves Roger Directie Algemeen Directeur: Raf Mertens Adjunct Algemeen Directeur: Jean-Pierre Closon Contact Federaal Kenniscentrum voor de Gezondheidszorg (KCE) Administratief Centrum Kruidtuin, Doorbuilding (10e verdieping) Kruidtuinlaan 55 B-1000 Brussel Belgium Tel: +32 [0] Fax: +32 [0] Web:

3 Kosteneffectiviteit van 10- en 13-valent geconjugeerde pneumokokkenvaccins bij kinderen KCE reports 155A PHILIPPE BEUTELS, ADRIAAN BLOMMAERT, GERMAINE HANQUET, JOKE BILCKE, NANCY THIRY, MARTINE SABBE, JAN VERHAEGEN, FRANK DE SMET, MICHAEL CALLENS, PIERRE VAN DAMME Federaal Kenniscentrum voor de Gezondheidszorg Centre fédéral d expertise des soins de santé 2011

4 KCE reports 155A Titel: Kosteneffectiviteit van 10- en 13-valent geconjugeerde pneumokokkenvaccins bij kinderen Auteurs: Philippe Beutels (Centre for Health Economics Research & Modelling Infectious Diseases (CHERMID), Vaccine & Infectious Disease Institute, University of Antwerp), Adriaan Blommaert (Centre for Health Economics Research & Modelling Infectious Diseases (CHERMID), Vaccine & Infectious Disease Institute, University of Antwerp), Germaine Hanquet (KCE), Joke Bilcke (Centre for Health Economics Research & Modelling Infectious Diseases (CHERMID), Vaccine & Infectious Disease Institute, University of Antwerp), Nancy Thiry (KCE), Martine Sabbe (Wetenschappelijk Instituut Volksgezondheid), Jan Verhaegen (Dept Microbiologie, UZ Leuven), Frank De Smet (Christelijke Mutualiteit), Michael Callens (Christelijke Mutualiteit), Pierre Van Damme (Centre for Evaluation of Vaccination, Vaccine & Infectious Disease Institute, University of Antwerp). Herzieners: Chris De Laet (KCE), Frank Hulstaert (KCE), France Vrijens (KCE), Jo Robays (KCE). Acknowledgements : We danken graag volgende personen voor het verstrekken van informatie: Anne Kongs (Vlaams Agentschap Zorg en Gezondheid), Murielle Deguerry en Peter Verduyckt (Observatoire de la santé et du social de Bruxelles-Capitale), Marie-Anne Goosse (Ministère de la Communauté Française). Externe experten: Johan Bots (Gemeenschappelijke Gemeenschapscommissie), Iris Deschutter (UZ Brussel), Pieter Neels (Federaal Agentschap voor Geneesmiddelen en Gezondheidsproducten), Willy Peetermans (UZ Leuven), Marijke Proesmans (UZ Leuven), Martine Sabbe (Wetenschappelijk Instituut Volksgezondheid), Toon Braeye (Wetenschappelijk Instituut Volksgezondheid), Beatrice Swennen (ULB), Geert Top (Vlaams Agentschap Zorg en Gezondheid), David Tuerlinckx (Cliniques Universitaires Mont Godinne), Yves Van Laethem (UMC Sint- Pieter), Stefaan Van Lierde (Regionaal Ziekenhuis Heilig Hart Tienen), Anne Vergison (Universitair KinderZiekenhuis Koningin Fabiola). Externe validatoren: Philippe De Wals (Université Laval, Canada), Daniel Levy-Bruhl (Institut de Veille Sanitaire, France), Steven Simoens (Universiteit Leuven, Belgïe). Belangen conflict: Willy Peetermans heeft een fonds ontvangen voor het uitvoeren van onderzoek en een reisvergoeding voor deelname aan verschillende symposiums. Béatrice Swennen heeft betalingen ontvangen om te spreken, alsmede een opleidingsvergoeding en reisvergoeding voor deelname aan een symposium. Yves Van Laethem heeft een fonds ontvangen voor het uitvoeren van onderzoek en heeft tevens honoraria ontvangen voor raadpleging vanwege de farmaceutische bedrijven GSK en Wyeth/Pfizer. Hij heeft bovendien ook reisvergoedingen ontvangen voor deelname aan verschillende symposiums. Iris Deschutter heeft een fonds ontvangen voor het uitvoeren van onderzoek en ook reisvergoedingen gekregen voor deelname aan verschillende symposiums vanwege GSK en Wyeth/Pfizer. David Tuerlinckx heeft honoraria ontvangen voor consultancy, alsook reisvergoedingen voor deelname aan verschillende symposiums vanwege GSK en Wyeth/Pfizer. Anne Vergison heeft betalingen ontvangen voor consultancy en te spreken voor GSK en Wyeth/Pfizer, alsook een fonds voor uitvoering van onderzoek. Jan Verhaegen is lid van de advisory boards voor Pfizer en GSK en ontving reisvergoeding voor deelname aan symposiums. In 2009, Martine Sabbe was externe consultant (niet-betaalde) voor Wyeth/Pfizer in verband met de studie «Pneumococcal vaccination in the elderly population. Pierre Van Damme is hoofdonderzoeker voor vaccin studies uitgevoerd voor de

5 Universiteit Antwerpen, waarvoor de Universiteit onderzoeksfondsen ontvangt van vaccin producenten. De vergoedingen ontvangen voor het geven van voordrachten over vaccins worden gestort op een educatief fonds van de Universiteit. De wetenschappelijke leerstoel Evidence-Based Vaccinology aan de Universiteit Antwerpen werd in de periode gefinancierd met een gift van Pfizer. Disclaimer : De externe experten werden geraadpleegd in verband met een (preliminaire) versie van het wetenschappelijk rapport. Tijdens de vergaderingen werd hun commentaar besproken. Ze waren geen medeauteur van het wetenschappelijk rapport en gingen niet noodzakelijk akkoord met de inhoud ervan. Nadien werd een (definitieve) versie voorgelegd aan de validatoren. De bevestiging van het rapport vloeide voort uit een consensus of uit een stemproces tussen de validatoren. De validatoren waren geen medeauteur van het wetenschappelijk rapport en gingen niet noodzakelijk akkoord met de inhoud ervan. Ten slotte werd dit rapport unaniem goedgekeurd door de Raad van Bestuur. - Uitsluitend de KCE is verantwoordelijk voor vergissingen of weglatingen die overblijven. De beleidsaanbevelingen vallen ook onder de volledige verantwoordelijkheid van de KCE. Layout: Ine Verhulst Brussel, 30 mei 2011 Studienr: Domein: Health Technology Assessment (HTA) MeSH: Vaccines, Conjugate; Pneumococcal vaccines; Infant; Pneumococcal Infections; Cost-Benefit Analysis NLM classificatie: WC 217 Taal: Nederlands, Engels Formaat: Adobe PDF (A4) Wettelijk depot: D/2011/10.273/19 Dit document is beschikbaar op de website van het Federaal Kenniscentrum voor de Gezondheidszorg KCE reports worden gepubliceerd onder een by/nc/nd Creative Commons Licence ( Hoe refereren naar dit document? Beutels P, Blommaert A, Hanquet G, Bilcke J, Thiry N, Sabbe M, Verhaegen J, De Smet F, Callens M, Van Damme P. Kosteneffectiviteit van 10- en 13-valent geconjugeerde pneumokokkenvaccins bij kinderen. Health Technology Assessment (HTA). Brussel: Federaal Kenniscentrum voor de Gezondheidszorg (KCE) Reports 155A. D/2011/10.273/19

6

7 KCE report 155A 10 en 13-valent geconjugeerde pneumokokkenvaccins i VOORWOORD Al draagt hij de naam Streptococcus pneumoniae, toch veroorzaakt de pneumokok (want zo wordt deze bacterie courant genoemd) ook nog een hele reeks andere infecties dan alleen maar longontstekingen. En daaronder niet van de minste: hersenvliesontsteking, bloedbaaninfectie, septische shock Daarnaast ook minder ernstige, maar veel frequentere infecties zoals middenoorontsteking. Vooral jonge kinderen, beneden de 2 jaar en oudere personen vormen de risicopopulatie. Er zijn talrijke verschillende varianten van de pneumokok in omloop men spreekt van serotypes zodat een doeltreffend vaccin in feite steeds een combinatie van verschillende vaccins moet zijn, die elk op een specifiek serotype zijn gericht. In 2006 publiceerde het KCE een eerste rapport over het toen beschikbare vaccin dat bescherming bood tegen 7 serotypes (PCV7). Ons advies was gematigd positief, maar we wezen op de vele onbekenden in de analyses, zodat de voorspellingen van het toekomstig nut een grote onzekerheidsmarge hadden. Nu, vijf jaar later, vraagt men vanuit de overheid om een uitspraak te doen over twee nieuwe vaccins, met respectievelijk 10 en 13 serotypes. Intussen heeft de Hoge Gezondheidsraad zich reeds voorzichtig gunstig uitgesproken ten voordele van het laatste vaccin. Dit lijkt a priori logisch: hoe meer serotypes, hoe beter de bescherming. Maar zo eenvoudig liggen de zaken niet. Zo blijkt in dit geval 13 niet gewoon 10 plus 3 te zijn. Daarnaast hebben we te maken met verschillende toedieningsschema s, verschillende prijzen en een heleboel onzekerheden. Wat zal de uiteindelijke prijs zijn die kan worden bedongen? Zijn de voorspelde positieve effecten van PCV7 bevestigd? Zullen er geen andere serotypes de kop opsteken? En zo kan men nog even doorgaan. Daarom zijn er in dit soort studies ook zelden zwart-wit conclusies. Zoveel mogelijk objectieve informatie aanreiken aan de beleidsmakers: dit is de ambitie van dit rapport, het zoveelste werkstuk dat voortkomt uit een samenwerking met de equipe van de Universiteit Antwerpen. Wij danken hun voor hiervoor, evenals de vele andere experts die ons met hun gewaardeerd advies hebben bijgestaan. Jean Pierre CLOSON Adjunct Algemeen Directeur Raf MERTENS Algemeen Directeur

8 ii 10 en 13-valent geconjugeerde pneumokokkenvaccins KCE report 155A INLEIDING Samenvatting van het rapport INFECTIE MET PNEUMOKOKKEN Streptococcus pneumoniae (of pneumococcus ) is een ziekteverwekkende bacterie die wereldwijd kinderen en volwassenen treft. Hij bestaat in meer dan 90 serotypes die niet alleen in structuur verschillen maar ook in de ziekte die ze kunnen veroorzaken en in leeftijdsgroepen die het meest getroffen worden. S. pneumoniae is de voornaamste oorzaak van meningitis, pneumonie en otitis media bij jonge kinderen, maar kan ook ernstige morbiditeit veroorzaken bij ouderen, bij wie de mortaliteit het hoogst is. De ernstigste aandoeningen veroorzaakt door S. pneumoniae zijn de invasieve pneumokokkenziekten (invasive pneumococcal disease, IPD) Deze invasieve vormen zijn meningitis en bacteriëmie, die kunnen leiden tot septische shock. De IPD treffen de uiterste leeftijdsgroepen, met name de jonge kinderen en de ouderen. De niet-invasieve ziekten omvatten hoofdzakelijk pneumonie en otitis media. Zij zijn doorgaans minder ernstig, maar komen aanzienlijk meer voor dan de IPD. Otitis media komt voornamelijk voor bij jonge kinderen. Pneumonie, daarentegen, komt voor in alle leeftijdsgroepen. De mortaliteit en de morbiditeit door IPD blijven ook vandaag hoog, ondanks een adequate toegang tot zorg en een behandeling met antibiotica. In verschillende delen van de wereld wordt de behandeling van een infectie met pneumokokken bemoeilijkt door de toenemende resistentie tegen penicilline en andere antibiotica. In België werd de jaarlijkse (2005) ziektelast te wijten aan Streptococcus pneumoniae tijdens de periode vóór vaccinatie geschat op 1403 gevallen van IPD waaronder 96 gevallen van meningitis en 500 bacteriëmieën. Streptococcus pneumoniae was verantwoordelijk voor 53 sterfgevallen waarvan 12 te wijten aan meningitis. Daarnaast zagen huisartsen op hun raadpleging ongeveer gevallen van otitis media en gevallen van pneumonie in Naar schatting 30-50% van deze gevallen waren toe te schrijven aan S. pneumoniae. PCV7 VACCINATIE In 2001 werd een eerste geconjugeerd pneumokokkenvaccin (PCV) goedgekeurd in Europa. Dit vaccin bevat antigenen van zeven serotypes (4, 6B, 9V, 14, 18C, 19F en 23F) die we verder vaccinserotypes noemen. In de Verenigde Staten leidde de introductie van PCV7 vaccinatie in het vaccinatieschema tot een spectaculaire afname van het aantal IPD gevallen. Bovendien werd een belangrijk indirect effect vastgesteld, met name de afname van IPD bij nietgevaccineerde groepen door een verminderde circulatie van pathogenen na de massieve vaccinatie van kinderen. Dit wordt herd immunity of groepsimmuniteit genoemd. PCV7 leidde echter ook tot een toename in IPD door serotypes die niet in het vaccin zitten (de zogenoemde niet-vaccinserotypes) ten gevolge van een vervangingseffect waarbij vaccinserotypes vervangen worden door niet-vaccinserotypes. In België waren deze 7 serotypes van PCV7 verantwoordelijk voor 72% van de IPD gevallen bij kinderen <2 jaar. PCV7 was aanbevolen in België maar het werd slechts in oktober 2004 beschikbaar omwille van een vaccintekort. In een eerste fase werd PCV7 aanbevolen voor kinderen <2 jaar, maar de kosten werden grotendeels door de ouders gedragen terwijl de meeste ziekenfondsen het geleidelijk aan gedeeltelijk terugbetaalden. In juni 2006 werd in een KCE studie de effectiviteit en de kosteneffectiviteit onderzocht van de vaccinatie met het PCV7 vaccin bij kinderen. Dit rapport concludeerde dat de kosteneffectiviteit van een veralgemeende vaccinatie van kinderen met PCV7 niet duidelijk was omwille van de onzekerheden geassocieerd met de effecten van groepsimmuniteit en van serotypevervanging. Het KCE rapport argumenteerde echter dat de onzekerheid op gebied van kosteneffectiviteit lager zou zijn als men gebruik maakte van het 2+1 schema (een eerste reeks van 2 doses gevolgd door 1 booster) in plaats van het 3+1 schema dat destijds werd aanbevolen.

9 KCE report 155A 10 en 13-valent geconjugeerde pneumokokkenvaccins iii In aansluiting op dit KCE rapport besloot de Interministeriële conferentie in juni 2006 om in België de veralgemeende PCV7 vaccinatie van kinderen te introduceren gebruik makend van een 2+1 schema, d.w.z. vaccinatie op de leeftijd van 2, 4 en 12 maanden. Deze beslissing leidde tot de gratis opname van het PCV7 vaccin in de regionale vaccinatieprogramma s vanaf januari 2007, met een inhaalvaccinatie voor kinderen tot 2 jaar oud. De opname van het PCV7 vaccin werd op respectievelijk 81 en 89% geschat voor het volledige schema in Wallonië en Vlaanderen in Ondanks de hoge vaccinatiegraad verminderde de algemene incidentie van IPD slechts matig na de PCV7 introductie in België (-37% tot -46% in kinderen jonger dan 2 jaar in 2008 vergeleken met ). Gegevens van IPD surveillance na vaccinatie toonden een hoge en snelle impact op de vaccinserotypes in alle pediatrische leeftijdsgroepen, maar ook een bijkomende stijging in de niet-vaccinserotypes. Dit werd waargenomen voor de serotypes 33F, 10A, 12F, 24F en in het bijzonder voor serotypes 1, 7F en 19A die in 2008 samen 55% van de IPD in België uitmaakten. Maar het serotype 19A was in België al toegenomen vóór PCV7 werd geïntroduceerd. De mate waarin deze toename te wijten is aan serotypevervanging is bijgevolg onduidelijk en andere factoren zoals natuurlijke temporele trends en het gebruik van antibiotica spelen waarschijnlijk ook een rol. Men verwachtte ook van PCV7 dat het doeltreffend zou zijn tegen acute otitis media (AOM) en tegen pneumonie. Maar de routine opvolging door Intego, een registratienetwerk van Vlaamse huisartsen, suggereerde geen bewijs van enige impact. Bovendien kon op basis van de beschikbare gegevens geen groepsimmuniteit vastgesteld worden, in tegenstelling tot de ervaring in de VS. Dit is echter consistent met de bevindingen in de meeste andere Europese landen. NIEUWE GECONJUGEERDE PNEUMOKOKKENVACCINS: PCV10 EN PCV13 Twee nieuwe geconjugeerde pneumokokkenvaccins kregen in 2009 de marktvergunning van de Europese commissie. Deze vaccins dekken zowel de 7 serotypes in het PCV7 vaccin als bijkomende serotypes die nu verantwoordelijk zijn voor een groot deel van de invasieve ziekten. Synflorix of PCV10 (GSK) is een 10-valent vaccin dat bijkomende antigenen bevat van de serotypes 1, 5 en 7F. De fabrikant beweert dat er een grote bescherming is tegen AOM te wijten aan pneumokokkenserotypes maar ook tegen AOM ten gevolge van niettypeerbare Haemophilus influenzae (NTHi) omdat het dragereiwit van het vaccin afgeleid is van Haemophilus influenzae. Er is eveneens discussie over het feit of PCV10 bescherming zou kunnen bieden tegen het serotype 19A door zogenaamde kruisbescherming die zou worden teweeggebracht door de immuunrespons tegen serotype 19F dat tot dezelfde groep behoort. Prevenar 13 of PCV13 (Pfizer) is een 13-valent vaccin dat bijkomende antigenen van de serotypes 1, 3, 5, 6A, 7F en 19A bevat. DOELSTELLINGEN VAN DE HUIDIGE STUDIE Omwille van de sterke toename van de serotypes 1, 5 en 7F die men momenteel in België observeert en de relatief beperkte impact van PCV7 vaccinatie lijkt het voor de hand liggend om over te schakelen van PCV7 naar de nieuwe PCV vaccins omdat ze beide de opkomende serotypes dekken. Het is echter moeilijk te bepalen welk nieuw vaccin te verkiezen is. PCV10 kan een grotere bescherming bieden tegen AOM, terwijl PCV13 meer serotypes dekt die IPD veroorzaken (zeker tegen het opkomende serotype 19A). De keuze tussen deze vaccins moet ook gemaakt worden in een context van onzekerheid met betrekking tot toekomstige serotypevervanging (aangezien andere opkomende serotypes niet worden gedekt door de nieuwe vaccins) en moet rekening houden met de kostprijs van de vaccins. Met het oog op de beschikbaarheid van deze nieuwe vaccins is dit rapport bedoeld om de incrementele effectiviteit en de kosteneffectiviteit te bepalen van de vervanging van PCV7 door PCV10 of PCV13 in België. Hierbij wordt rekening gehouden met effecten van groepsimmuniteit en serotypevervanging.

10 ii 10 en 13-valent geconjugeerde pneumokokkenvaccins KCE report 155A BESCHERMING DOOR DE NIEUWE GECONJUGEERDE PNEUMOKOKKENVACCINS PCV10 PCV13 Verder bouwend op het vorige KCE rapport startte het literatuuronderzoek om dit hoofdstuk te documenteren op 1 januari 2006 en duurde tot 1 maart Er bestaan slechts beperkte gegevens over de klinische werkzaamheid op basis van klinische studies voor PCV10 en PCV13. Gebruik makend van de correlaten van bescherming voor de werkzaamheid van het vaccin waren de studies eerder geconcentreerd op merkers die aanduiden dat een immuunrespons werd opgewekt (per serotype), uitgedrukt door de concentraties aan antilichamen en opsonofagocytische activiteit (OPA). De OPA meet de activiteit van de antilichamen en hun vermogen om de pneumokokken te elimineren en wordt bijgevolg beschouwd als een betere merker voor de klinische bescherming. Er waren geen gegevens over de werkzaamheid van deze nieuwe vaccins. Er werd aangetoond dat PCV10 antilichaam opwekt tegen alle pneumokokkenserotypes in het vaccin. Het werd ook aangetoond dat PCV10 niet inferieur was aan PCV7 voor alle 7 serotypes in een 3+1 schema. De immuunrespons tegen de bijkomende serotypes was hoog, hoewel de OPA respons op serotypes 1 en 5 lager was dan op de andere 7 serotypes van PCV7. Verder waren er, op basis van de immuunresponsen, enkele elementen die erop wijzen dat PCV10 bescherming induceert tegen het opkomende serotype 19A. Eén klinische studie maakte gebruik van het PCV10 precursor vaccin (PCV11), dat gelijkaardig is aan PVC10 maar in tegenstelling tot PVC10 ook het serotype 3 antigen bevat, om de klinische effectiviteit van PCV10 tegen AOM te bepalen. In deze studie werd aangetoond dat de bescherming tegen AOM 34% was. De studie suggereerde dat het vaccin niet alleen tegen pneumokokkenserotypes bescherming biedt maar ook tegen niet-typeerbare Haemophilus influenzae (NTHi). De studies met PCV13 toonden aan dat de antilichaamresponsen en OPA-niveaus van PCV13 niet inferieur waren aan PCV7 voor de 7 serotypes. Er werd ook aangetoond dat er voldoende immuunresponsen waren voor de bijkomende serotypes behalve voor serotype 3 dat een lagere immuunrespons vertoonde. Geen enkele studie evalueerde de klinische werkzaamheid van PCV13. ECONOMISCHE EVALUATIE VAN PCV10 EN PCV13 LITERATUURSTUDIE Deze studie was beperkt tot evaluaties van PCV10 en/of PCV13 gepubliceerd tussen 1 januari 2006 (de zoeklimiet van het vorige KCE rapport) en 1 maart Met uitzondering van één studie bevatte geen van de 8 geïdentificeerde gepubliceerde economische evaluaties de effecten van groepsimmuniteit en serotypevervanging om de kosteneffectiviteit van de pneumokokkenvaccins te bepalen. De meeste evaluaties hebben de neiging om te besluiten dat PCV13 meer kosteneffectief is dan PCV10. Slechts één studie vermeldt het tegengestelde, met name dat PCV10 meer kosteneffectief is dan PCV13, als men vertrekt van de hypothese dat er een effect is van PCV10 op AOM door NTHi. De meeste economische evaluaties vertonen grote gebreken. Ten eerste werd geen rekening gehouden met de effecten van serotypevervanging en groepsimmuniteit. Ten tweede werden geen gevoeligheidsanalyses uitgevoerd op de kostprijs van de vaccins, die toch een zeer invloedrijke parameter is. Ten derde werd de impact van verschillende hypothesen voor het bepalen van de relatieve werkzaamheid van de vaccins niet onderzocht, terwijl deze werkzaamheid toch cruciaal is omdat er nog steeds bijna geen klinische gegevens zijn over PCV10 en PCV13.

11 KCE report 155A 10 en 13-valent geconjugeerde pneumokokkenvaccins v In de economische evaluatie die in het kader van huidig project werd opgesteld, maken we projecties van de impact van PCV10 en PCV13 waarbij deze aspecten wel in aanmerking worden genomen. KOSTEN-NUTSANALYSE VAN PCV10 EN PCV13 IN BELGIË Beschrijving van het model en veronderstellingen Er is een simulatiemodel ontwikkeld dat de incidentie en gevolgen van infecties met pneumokokken nabootst in cohorten van gevaccineerde kinderen en in de algemene bevolking. Om dit model te parameteriseren werden zowel Belgische gegevensbronnen als de internationale literatuur gebruikt. Dit model integreert de effecten van groepsimmuniteit en van serotypevervanging en houdt ook rekening met de mate waarin PCV10 bijkomende bescherming biedt tegen AOM versus PCV13 en met de mate waarin PCV10 een zekere bescherming biedt tegen serotype 19A. De serotype-specifieke vaccin werkzaamheid tegen IPD werd onrechtstreeks afgeleid uit immunologische gegevens (antilichamen en OPA metingen) en werd aangepast op basis van observationele gegevens over de werkzaamheid van het PCV7 vaccin. Het effect van serotypevervanging werd gemodelleerd als een reductie in serotypedekking voor IPD en als een reductie in de vaccin werkzaamheid voor AOM en pneumonie (aangezien er geen serotype-specifieke gegevens over deze aandoeningen beschikbaar zijn). Deze benadering impliceert dat eenzelfde parameter voor serotypevervanging zal resulteren in méér vervanging met PCV13 dan met PCV10. In scenarioanalysen worden de hypothesen over groepsimmuniteit, serotypevervanging, mogelijke bijkomende bescherming van PCV10 tegen AOM en 19A, en klinische werkzaamheid uitgebreid onderzocht. De rechtstreekse gezondheidskosten werden geraamd op basis van een intensieve nationale face-to-face enquête. De kostprijs voor PCV7, PCV10 en PCV13 werd vastgelegd volgens hun huidige prijs in de apotheek, met name 66,15, 70,44 en 74,55 per dosis, maar deze waarden werden uitgebreid gevarieerd in het simulatiemodel (met inbegrip van een hypothese van gelijkheid van prijzen). Rekening houden met het in voege zijnde vaccinatieschema voor kinderen en het feit dat PCV10 momenteel enkel vergund is voor het 3+1 schema, concentreren we ons op de volgende vaccinatie opties: (1) de huidige situatie met PCV7 vaccinatie volgens een 2+1 schema; (2) PCV13 vaccinatie volgens een 2+1 of een 3+1 schema; (3) PCV10 vaccinatie volgens een 3+1 schema. Een 2+1 schema werd ook onderzocht voor PCV10, in afwachting van mogelijke veranderingen van het toegelaten schema. Opties 2 en 3 worden vergeleken met optie 1 en ook incrementeel met elkaar. Resultaten - PCV10 en PCV13 versus PCV7 Door gebruik te maken van verschillende veronderstellingen over de werkzaamheidsmetingen van het vaccin en over de effecten van groepsimmuniteit en serotypevervanging toonden de resultaten constant dat beide nieuwe vaccins zeer waarschijnlijk kostenbesparend of kosteneffectief zijn in vergelijking met PCV7, zelfs aan hun huidige prijs in de apotheek. Vergeleken met PCV7 en zonder het effect van groepsimmuniteit vermeden de nieuwe PCV10 en PCV13 vaccins, volgens alle scenario s, 113 tot 118 IPDs, 181 tot 236 pneumonieën, 587 tot otitis media en iets minder dan 2 overlijdens. De incrementele kosteneffectiviteitsratio van PCV10 (3+1 schema) en PCV13 (2+1 schema) ten opzichte van PCV7 varieerde tussen dominantie (d.w.z. dat de nieuwe vaccins werkzamer en goedkoper zijn dan PCV7), en per gewonnen QALY (qualityadjusted life-year, gezond levensjaar). Deze waarden waren gunstiger wanneer groepsimmuniteitseffecten werden verondersteld. We zouden ons echter de vraag kunnen stellen of PCV7 vaccinatie nog kan worden beschouwd als een kosteneffectieve interventie in België (en dus de aangewezen comparator voor PCV10 en PCV13) gezien de geobserveerde serotypevervanging. Dit was echter geen doelstelling van deze studie.

12 ii 10 en 13-valent geconjugeerde pneumokokkenvaccins KCE report 155A De resultaten toonden verder aan dat het 3+1 schema waarschijnlijk geen goede optie is in vergelijking met het 2+1 schema als de prijs van het vaccin per dosis constant blijft tussen de schema s. De vergelijking tussen de 3+1 en 2+1 schema s is momenteel echter enkel relevant voor PCV13, dat een vergunning heeft voor beide schema s (en PCV10 enkel voor het 3+1 schema). Resultaten PCV10 versus PCV13 De voorkeur tussen PCV10 en PCV13 verschilde naargelang het belang dat de beleidsmaker hecht aan de preventie van ernstige ziekten (IPD en pneumonie) bij weinig individuen of aan de preventie van milde ziekten (otitis media) bij veel kinderen. Doorgaans, bij gelijke vaccinprijzen, werd de voorkeur gegeven aan PCV13 in plaats van PCV10 als men in acht neemt dat deze vaccins enkel een impact hebben op ernstige ziekten (d.w.z. exclusief otitis media). In dat geval vermijdt PCV13 meer behandelingskosten en wint meer QALY s dan PCV10 (Figuur 1). Daar staat tegenover dat men, rekening houdend met het grote aantal gevallen van otitis media, meende dat PCV10 meer behandelingskosten zou vermijden dan PCV13 als de impact op otitis media wordt ingecalculeerd. Maar PCV10 won minder QALY s dan PCV13. Figuur 1: Mediane rechtstreekse besparingen ( ) in behandelingskosten en mediane gewonnen QALY s volgens de klinische manifestatie voor verschillende beslissende standpunten over PCV10 (3+1) of PCV13 (2+1). Scenario inclusief serotypevervanging, kruisbescherming voor serotype 19A van PCV10 en exclusief groepsimmuniteit. NTHi: Niet-typeerbare haemophilus influenzae; IPD: invasieve ziekte door pneumokokken. Welke opties ook in aanmerking worden genomen, d.w.z. inclusief of exclusief groepsimmuniteit en de impact van PCV10 tegen niet-typeerbare Haemophilus influenzae, de impact van variaties in de vaccinprijs is het sterkste en gaat altijd in dezelfde richting: de voorkeur gaat naar PCV10 (3+1) aan de huidige publieksprijs en naar PCV13 (2+1) bij gelijke prijzen. Door gebruik te maken van publieksprijzen, heeft men gevonden dat PCV10 (3+1) meer kosteneffectief was dan PCV13 in 96% van de simulaties (inclusief werkzaamheid tegen niet-typeerbare Haemophilus influenzae otitis media en exclusief groepsimmuniteit), in 88% van de simulaties (exclusief effectiviteit tegen NTHi otitis media en exclusief groepsimmuniteit) en in 51% van de simulaties (exclusief werkzaamheid tegen NTHi otitis media en inclusief groepsimmuniteit. Door gebruik te maken van dezelfde prijs per dosis voor PCV13 en voor PCV10 bedroegen deze percentages respectievelijk 65%, 23% en 4%; waardoor PCV13 de meest gewenste optie wordt in vergelijking met PCV10 (Tabel 1).

13 KCE report 155A 10 en 13-valent geconjugeerde pneumokokkenvaccins vii Groepsimmuniteit voor IPD Tabel 1: Waarschijnlijkheid dat PCV10 (3+1) dominant is of een incrementele kosteneffectiviteitsratio < per QALY heeft ten opzichte van PCV13 (2+1) voor verschillende veronderstellingen over de prijzen van de vaccins, over de groepsimmuniteit en over de werkzaamheid tegen Niet-typeerbare Haemophilus influenzae (NTHi) otitis media. In elk scenario wordt kruisbescherming voor 19A van PCV10 en uitgebreide serotypevervanging verondersteld. HUIDIGE PUBLIEKSPRIJS GELIJKHEID VAN PRIJS (IN DE APOTHEEK) Effectiviteit tegen NTHi otitis media Ja Nee Ja Nee Ja - 51% 28% 4% Nee 96% 88% 65% 23% De resultaten waren ook uiterst gevoelig voor veronderstellingen over serotypevervanging. In de veronderstelling dat er geen serotypevervanging zou zijn en geen hogere impact op AOM van PCV10 in vergelijking met PCV13, maar wel kruisbescherming voor serotype 19A bij PCV10, dan is PCV13 te verkiezen. In het scenario waarbij PCV10 een hogere impact heeft op AOM dan PCV13, waar groepsimmuniteit zich niet voordoet (zoals vastgesteld met PCV7 in België en in verschillende andere Europese landen, in tegenstelling tot de VS) en er kruisbescherming is voor 19A bij PCV10, blijft PCV13 te verkiezen als de effecten van serotypevervanging klein zijn. Met toenemende serotypevervanging in IPD alleen neemt het voordeel van PCV13 ten opzichte van PCV10 af, vooral als PCV10 bijkomende werkzaamheid induceert tegen AOM. Als PCV10 10% meer werkzaamheid oplevert tegen AOM is het voordeel voor PCV10. Dit onverwachte resultaat is te wijten aan de methode die we hebben gebruikt om serotypevervanging te simuleren in ons model. Als we de veronderstelde kruisbescherming voor serotype 19A bij PCV10 echter uitsluiten, zijn de resultaten opnieuw gunstiger voor PCV13. DISCUSSIE EN BESLUIT We hebben een populatiesimulatiemodel ontwikkeld om de incrementele werkzaamheid en kosteneffectiviteit te bepalen als PCV7 wordt vervangen door PCV10 of PCV13 in België. Uit de resultaten met betrekking tot de werkzaamheid van de vaccins en de ziektelast van pneumokokkengerelateerde aandoeningen, blijkt dat beide vaccins een klinisch voordeel hebben ten opzichte van de huidige PCV7 vaccinatie. Dit voordeel is nog relevanter door de recente toename van de bijkomende serotypes die PCV10 en PCV13 bevatten, hetgeen verklaart waarom PCV7 niet helemaal voldeed aan de verwachtingen van de vorige kosteneffectiviteitsanalyse. Het is echter veel minder duidelijk welk van beide vaccins moet worden gekozen omdat de verschillen tussen de twee vaccins niet voor de hand liggen en er nog veel onzekerheden zijn hieromtrent. In ieder geval zal de keuze van het vaccin in sterke mate worden bepaald door de prijzen waaraan beide vaccins in grote hoeveelheid worden aangeboden. Bij de huidige publieksprijs is PCV13 (2+1) minder kosteneffectief dan PCV10 (3+1) (waarbij kruisbescherming voor 19A wordt verondersteld). Maar na een aanbestedingsprocedure zouden de prijsverschillen tussen beide vaccins kunnen veranderen en zou PCV13 een kosteneffectievere optie kunnen worden dan PCV10, afhankelijk van de gekozen scenario s. Daarnaast zal de keuze tussen deze twee vaccins ook worden bepaald door het relatieve belang dat men zal hechten aan de preventie van milde aandoeningen bij veel kinderen (met name AOM) versus de preventie van zeer ernstige aandoeningen in minder frequente gevallen (met name IPD).

14 ii 10 en 13-valent geconjugeerde pneumokokkenvaccins KCE report 155A Tenslotte hing de keuze tussen de vaccins in belangrijke mate af van de verwachte serotypevervangingseffecten en de potentiële bescherming van PCV10 tegen NTHi AOM en tegen 19A, effecten die nog moeten worden aangetoond. In de veronderstelling dat de bescherming van PCV10 tegen AOM hoger is dan die van PCV13, is PCV13 te verkiezen zolang het serotypevervangingseffect laag blijft. Met hogere niveaus van serotypevervanging en bescherming van PCV10 tegen NTHi AOM wordt PCV13 minder wenselijk ten voordele van PCV10. Het uitsluiten van het eerder veronderstelde beschermingseffect van PCV10 op serotype 19A leidde opnieuw tot resultaten ten voordele van PCV13. Hoewel we de evolutie van Streptococcus pneumoniae zo accuraat mogelijk hebben proberen te simuleren met de beste kennis tot nu toe, heeft ons model een aantal belangrijke beperkingen. Ten eerste hebben we de gevolgen van serotypevervanging op individuele serotypes en van variaties in pathogeniciteit over serotypes heen niet meegenomen in het model. Ten tweede hebben we geen rekening gehouden met resistentie tegen antibiotica (een probleem waarover we ons zorgen maken in geval van serotype 19A). Ten derde hebben we in het model geen rekening gehouden met de gestage toename van 19A, en benadeelt onze methode om serotypevervanging te modelleren één vaccin meer dan het andere. Tenslotte hebben we de transmissiedynamieken van pneumokokkenserotypes niet gemodelleerd. We kunnen besluiten dat de resultaten van ons model pleiten voor een vervanging van het huidige PCV7 vaccin door het nieuwe PCV10 of PCV13 vaccin. De resultaten tonen ook duidelijk aan dat, voor beide vaccins, een 2+1 schema wenselijker is dan een 3+1 schema. De keuze tussen de PCV10 en PCV13 vaccins is echter niet vanzelfsprekend. De analyse en de interpretatie van de resultaten met betrekking tot de kosteneffectiviteit van ons model worden bemoeilijkt door de onzekere aard van toekomstige effecten van serotypevervanging en groepsimmuniteit en door de veronderstelde bescherming van PCV10 tegen serotype 19A en het NTHi AOM-effect van PCV10. Het laten variëren van deze veronderstellingen binnen redelijke grenzen maakte afwisselend PCV13 en PCV10 het te prefereren vaccin. Het effect van serotypevervanging en de potentiële bijkomende PCV10 bescherming tegen AOM en het serotype 19A waren de parameters met de grootste impact op de resultaten. Het is duidelijk dat de keuze tussen beide vaccins zal afhangen van de voorkeur van de beleidsmaker om te opteren voor de preventie van minder frequente ernstige IPD gevallen alleen of ook voor de preventie van frequente gevallen van AOM. De prijs waartegen beide vaccins zullen worden aangeboden zal eveneens een cruciale rol spelen, wat het belang benadrukt van een aanbestedingsprocedure om gunstige prijzen te verkrijgen.

15 KCE report 155A 10 en 13-valent geconjugeerde pneumokokkenvaccins ix AANBEVELINGEN 1 Gezien de waargenomen stijging van IPD veroorzaakt door serotypes die niet gedekt worden door het huidige PCV7 vaccin, is het vanuit klinisch oogpunt gewettigd over te schakelen naar de nieuwe PCV vaccins (PCV10 of PCV13). Het is aanbevolen om te opteren voor een 2+1 toedieningsschema indien beschikbaar; dit schema is kosteneffectiever dan een 3+1 schema. Momenteel heeft alleen PCV13 reeds een vergunning voor beide schema s en is PCV10 tot nu toe enkel vergund voor het 3+1 schema. De keuze tussen PVC10 en PVC13 is niet duidelijk omdat de verschillen tussen de twee vaccins niet voor de hand liggen en er nog veel onzekerheden zijn hieromtrent. De keuze zou door de volgende elementen kunnen beïnvloed worden: Het standpunt van de beleidsmaker. Als deze vooral streeft naar het voorkómen van ernstige ziektegevallen, dan is PCV13 de meest wenselijke optie. Als hij ernaar streeft om eveneens de vele gevallen van milde ziekte (waaronder AOM) te voorkómen, dan lijkt de voorkeur naar PCV10 te gaan, maar met een belangrijke graad van onzekerheid, wat dat betreft. De gehanteerde hypothesen. Het serotype-vervangings-effect en de potentiële bescherming van PCV10 tegen niet-typeerbare Haemophilus influenzae AOM en tegen serotype 19A zijn de meest invloedrijke parameters in het model. Het variëren van deze veronderstellingen binnen redelijke grenzen maakt afwisselend PCV13 of PCV10 het te verkiezen vaccin. De prijs van de vaccins. Aan de huidige publieksprijzen is PCV10 (3+1) waarschijnlijk kosteneffectiever. Bij gelijke prijzen voor beide vaccins is PCV13 (2+1) waarschijnlijk kosteneffectiever. Het KCE beveelt aan om een aanbestedingsprocedure uit te schrijven om gunstiger prijzen te bekomen; de prijs zou dan het bepalend element kunnen worden in het beslissingsproces. ONDERZOEKSAGENDA De mechanismen van groepsimmuniteit en serotypevervanging blijven onduidelijk en onvoorspelbaar. Wanneer deze mechanismen op een onverwachte manier zouden evolueren in de toekomst, kan het zijn dat onze aanbevelingen achterhaald worden. De evolutie van deze parameters zou dus moeten opgevolgd worden om de geldigheid van onze analyses te herevalueren. De werkzaamheid van PCV10 tegen niet-typeerbare Haemophilus influenzae AOM en de kruisbescherming tegen serotype 19A moet nog worden aangetoond. Er is nood aan een verdere verbetering van de simulatiemodellen voor Streptococcus pneumoniae door inclusie van de gevolgen van serotypevervanging op individuele serotypes, door rekening te houden met resistentie voor antibiotica en door directe modellering van de transmissiedynamiek van de verschillende pneumokokkenserotypes. 1 Alleen het KCE is verantwoordelijk voor de aanbevelingen aan de overheid

16

17 KCE Reports 155 Childhood pneumococcal vaccination 1 Scientific summary Table of contents LIST OF TABLES... 3 LIST OF FIGURES... 5 GLOSSARY BACKGROUND THE SEVEN-VALENT CONJUGATE PNEUMOCOCCAL VACCINE TWO NEW CONJUGATE PNEUMOCOCCAL VACCINES OBJECTIVES AND RESEARCH QUESTIONS LITERATURE SEARCHES AND METHODS IMMUNOGENICITY, EFFICACY AND EFFECTIVENESS OF CURRENT PNEUMOCOCCAL CONJUGATE VACCINES SEVEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV7) Clinical trials efficacy of PCV Post-licensure studies effectiveness of PCV Pathogenicity of serotypes TEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV10) Clinical trials immunogenicity of PCV Clinical trials efficacy of PCV THIRTEEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV13) Clinical trials immunogenicity of PCV DISEASE BURDEN IN BELGIUM INVASIVE PNEUMOCOCCAL DISEASE (IPD) ACUTE OTITIS MEDIA (AOM) AND COMMUNITY ACQUIRED PNEUMONIAE (CAP) HOSPITALISATIONS Meningitis hospitalisations Bacteremia and septicaemia hospitalisations Pneumonia hospitalisations DEATHS Invasive pneumococcal diseases (IPD) deaths (Pneumococcal) pneumonia deaths Pneumococcal meningitis deaths Pneumococcal septicaemia deaths ECONOMIC EVALUATIONS OF PCV10 AND PCV MODELS STRUCTURE MAIN ASSUMPTIONS AND RESULTS MATHEMATICAL MODELS OF PNEUMOCOCCAL TRANSMISSION COST-UTILITY ANALYSIS OF PCV10 AND PCV13 IN BELGIUM STUDY DESIGN General Vaccination options Mathematical model structure MODEL INPUT DATA Epidemiological parameters and transition probabilities Vaccine efficacy estimates Direct costs Health-Related Quality of Life RESULTS Cost-effectiveness acceptability curves Incremental costs, outcomes and cost-effectiveness ratios Further scenario analyses Influence of vaccine price and of inclusion or exclusion of AOM Influence of serotype replacement and herd immunity assumptions... 79

18 2 Childhood pneumococcal vaccination KCE reports Joint influence of vaccine price, expected serotype replacement and of the additional effectiveness of PCV10 versus PCV13 against otitis media Budget impact analysis CONCLUSION APPENDICES: ADDITIONAL COST-EFFECTIVENESS ACCEPTABILITY CURVES REFERENCES... 88

19 KCE Reports 155 Childhood pneumococcal vaccination 3 LIST OF TABLES Table 1. Comparison of case-control studies in Quebec and the US: % effectiveness of PCV7 against invasive pneumococcal disease (95% CI) Table 2. Basic characteristics of universal PCV7 programmes in a selection of European countries and Quebec Table 3. Clinical trials used to document immunogenicity of PCV Table 4. Clinical trials used to document immunogenicity of PCV Table 5. Estimated annual disease burden of pneumococcal infections pre-vaccination (2005) in Belgium, all ages pre-vaccination Table 6. Number and distribution of serotypes causing invasive pneumococcal disease in people under 16 years of age, according to clinical diagnoses (Belgium, 2009) Table 7. Distribution and invasive capacity of serotypes causing invasive pneumococcal disease in people under 16 years of age, according to clinical diagnoses (Belgium, 2009) Table 8. Age dependent frequency of clinical diagnoses associated with IPD cases under age 16 years (Belgium, 2009) Table 9. Deaths following IPD in children < 16 years old Table 10. Serotype by age group in children < 16 years who died following IPD Table 11. Deaths in Flanders with immediate or underlying cause recorded as pneumococcal meningitis Table 12. Deaths in Flanders with immediate or underlying cause recorded as pneumococcal septicemia Table 13. Main assumptions and results of economic evaluations on PCV10 or PCV Table 14a. Model input data related to population and disease burden Table 15. PCV10 immunogenicity trial data, produced after the infant primary course Table 16. PCV10 input data scaled to serotype-specific vaccine efficacy(%) based on US case-control study, and adjusted for polysorbate 80 based on trial 009 and inconsistencies in 2 priming versus 3 priming dose responses Table 17. PCV13 immunogenicity trial data, produced after the infant primary course Table 18. PCV13 immunogenicity input data scaled to serotype-specific vaccine efficacy (%) based on US case-control study Table 19. PCV13 input data scaled to serotype-specific vaccine efficacy (%) based on US case-control study, and adjusted for polysorbate 80 and inconsistencies in 2 priming versus 3 priming dose point estimate responses Table 20. Vaccine efficacy estimates used as input in the model Table 21. Assumed herd immunity, expressed as a reduction in cases at each age Table 22. Estimated median costs and effects (and 95% interval) of pneumococcal conjugate vaccination with PCV10 or PCV13 in a 2+1 schedule versus the current situation (PCV7 in a 2+1 schedule), using current public pharmacy vaccine prices, a 5 year time span for infections to accrue, a wide ranging uniform distribution of serotype replacement, and excluding herd immunity effects (results of 1000 model iterations) Table 23. Estimated median costs and effects (and 95% interval) of pneumococcal conjugate vaccination with PCV10 or PCV13 in a 3+1 schedule versus the current situation (PCV7 in a 2+1 schedule), using current public pharmacy vaccine prices, a 5 year time span for infections to accrue, a wide ranging uniform distribution of serotype replacement, and excluding herd immunity effects (results of 1000 model iterations) Table 24. Estimated median costs and effects (and 95% interval) of pneumococcal conjugate vaccination with PCV10 or PCV13 in a 3+1 schedule versus the same vaccine in a 2+1 schedule, using current public pharmacy vaccine prices, a 5 year time span for infections to accrue, a wide ranging uniform distribution of serotype replacement, and assuming no or equal herd immunity effects (results of 1000 model iterations) Table 25. Estimated median costs and effects (and 95% interval) of pneumococcal conjugate vaccination with PCV10 or PCV13 in a 2+1 schedule versus the current situation (PCV7 in a 2+1 schedule), using current public pharmacy vaccine prices, a 5 year time span for infections to accrue, a wide ranging uniform distribution of serotype replacement, and including herd immunity effects (results of 1000 model iterations) Table 26. Estimated median costs and effects (and 95% interval) of pneumococcal conjugate vaccination with PCV10 or PCV13 in a 3+1 schedule versus the current situation (PCV7 in a 2+1 schedule), using current public pharmacy vaccine prices, a 5 year time span for infections to

20 4 Childhood pneumococcal vaccination KCE reports 155 accrue, a wide ranging uniform distribution of serotype replacement, and including herd immunity effects (results of 1000 model iterations) Table 27. Additional scenario analyses at price parity (based on 1000 simulations at each row). Median direct costs ( ) per QALY gained (5 th percentile, 95 th percentile) Table 28. Evolution of mean additional annual health care costs ( ) over the first 5 years, assuming both vaccines would be purchased at a higher price than PCV7 (price PCV13 = current pharmacy price of PCV10)... 82

21 KCE Reports 155 Childhood pneumococcal vaccination 5 LIST OF FIGURES Figure 1. Literature search results for publications of interest in Medline(Pubmed), Web of Science and Scopus (1 st January st March 2011) Figure 2. Literature search results by vaccine formulation for publications potentially reporting on trials, retrieved through Medline(Pubmed), Web of Science and Scopus (1st January st March 2011) Figure 3. Halloran diagramme on vaccine effectiveness Figure 4. Incidence of invasive pneumococcal disease by serotype group in children <2 years of age and linear regression fitted on the pre-vaccination period (dotted lines) Figure 5. Incidence of invasive pneumococcal disease due to serotypes 1, 7F, 19A and 33F by age group Figure 6. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2007) Figure 7. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2008) Figure 8. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2009) Figure 9. Evolution of the incidence of GP consultations per 100,000 population due to pneumonia and acute otitis media, in Flanders Figure 10. Evolution of the age-specific incidence of GP consultations per 100,000 population due to acute otitis media and suspected pneumonia, in Flanders Figure 11. Incidence per 100,000 population of hospitalisations with meningitis (all causes) as diagnosis Figure 12. Incidence per 100,000 population of hospitalisations for patients diagnosed with bacteremia and septicaemia, all causes ( ) Figure 13. Deaths with pneumonia and pneumococcal pneumonia as the immediate or underlying cause of death (plotted on left and right hand axis) Figure 14. Basic structure of the static cohort model Figure 15. Cost-effectiveness acceptability curves ( ) for PCV10 (3+1 schedule) versus current situation (PCV7, 2+1 schedule), current public vaccine prices and a time span of 5 years, upper panel: no herd immunity, lower panel: with herd immunity Figure 16. Cost-effectiveness acceptability curves ( ) for PCV13 (2+1 schedule) versus current situation (PCV7, 2+1 schedule), current public vaccine prices and a time span of 5 years, upper panel: no herd immunity, lower panel: with herd immunity Figure 17. Cost-effectiveness acceptability curves ( ) for PCV13 (3+1 schedule) versus current situation (PCV7, 2+1 schedule), current public vaccine prices and a time span of 5 years, upper panel: no herd immunity, lower panel: with herd immunity Figure 18. Median direct savings ( ) in treatment costs, according to clinical manifestation for different decisional viewpoints on PCV10 or PCV13, including wide ranging uncertainty on serotype replacement on all clinical manifestations Figure 19. Median QALY gains according to clinical manifestation for different decisional viewpoints on PCV10 or PCV13, including wide ranging uncertainty on serotype replacement on all clinical manifestations Figure 20. Influence of price parity and of inclusion or exclusion of AOM Cost-effectiveness acceptability curves ( ) for a 2+1 schedule with PCV13 or a 3+1 schedule with PCV10 versus the current situation, assuming no herd immunity effects, a wide ranging distribution of serotype replacement, and a time span of 5 years Figure 21. Ratio of additional vaccine price per dose of PCV13 in a 3+1 schedule versus PCV10 in a 2+1 schedule at which both vaccines are equally cost-effective (at 30,000 per QALY gained), in relation to expected serotype replacement and the additional effectiveness of PCV10 versus PCV13 against otitis media. Upper panel: without herd immunity for IPD; lower panel: with herd immunity for IPD... 81

22 6 Childhood pneumococcal vaccination KCE reports 155 GLOSSARY ABBREVIATION ALOS AOM CAAP CAP CFR CM CPI DALY DTP EMA GDP GP HBV Hib IPD IPH IPV MMR MMRV MVTI NIS NTHI OM OPA OPV PCV QALY RCM-MKG RCT STR V VT DEFINITION Average length of stay Acute otitis media Community-acquired alveolar pneumonia Community-acquired pneumonia Case fatality ratio Christian Mutualities Consumer Price Index Disability-adjusted life-year Diphtheria Tetanus Pertussis European Medicine Agency Gross domestic product General practitioner Hepatitis B vaccine Haemophilus influenzae type b Invasive pneumococcal disease Scientific Institute of Public Health (Institut Scientifique de Santé Publique Wetenschappelijk Instituut of Volksgezondheid) Inactivated polio vaccine Measles mumps rubella Measles mumps rubella varicella Myringotomy with ventilation tube insertion National Institute for Statistics Non typable haemophilus influenza Otitis media Opsonophagocytic activity Oral polio vaccine Pneumococcal conjugate vaccine Quality-adjusted life-year Résumé Clinique Minimum Minimale Klinische Gegevens Randomized controlled trial Serotype replacement Varicella Vaccine serotypes

23 KCE Reports 155 Childhood pneumococcal vaccination 7 1 BACKGROUND Streptococcus pneumoniae (or pneumococcus ) is a bacterial pathogen that affects children and adults worldwide. It consists of more than 90 serotypes, which may have varying clinical consequences, depending on the serotype. It is a major cause of illness in children, especially those under the age of 2 years. S. pneumoniae can cause a wide spectrum of diseases. On the one hand, invasive diseases (IPD, defined as the isolation of pneumococcus in a normally sterile fluid) include meningitis, bacteraemia (presenting with or without focus of infection), bacteraemic pneumonia, sepsis, arthritis and peritonitis. On the other hand, non-invasive diseases mainly comprise lower respiratory tract infections (including non-bacteraemic pneumonia) and upper respiratory tract infections (including otitis media and sinusitis). In children, S. pneumoniae is one of the leading causes of meningitis, pneumonia and otitis media. Mortality and morbidity remain high despite appropriate access to care and antibiotic treatment. In various areas of the world, treatment of pneumococcal diseases is aggravated by the emergence of pneumococcal strains resistant to penicillin and other antibiotics THE SEVEN-VALENT CONJUGATE PNEUMOCOCCAL VACCINE A first pneumococcal conjugate vaccine (PCV) has been introduced in the US in 2000 and contains capsular polysaccharide antigens of seven serotypes (4, 6B, 9V, 14, 18C, 19F and 23F), each of them conjugated to a carrier protein. These 7 types were responsible for 80-90% of IPD in children less than 5 years of age in the USA, 50-60% in Europe, Latin America and Africa and 30-40% in Asia. 3 The seven-valent pneumococcal conjugate vaccine (denoted as PCV7 henceforth) was licensed in 2001 in Europe but became available in Belgium only in October 2004 due to vaccine shortage. Initial vaccination uptake was low and directed exclusively at groups at higher risk for complications, as it was initially not funded for the other groups. From 2005 on PCV7 vaccination uptake increased based on individual vaccine purchase and initiatives from private health insurers in Belgium to co-pay PCV7 for children. A previous KCE report by Beutels et al (June 2006) 4 examined the pre-vaccination disease burden and the potential effectiveness and cost-effectiveness of childhood vaccination using Prevenar the then licensed PCV7 vaccine. This KCE report concluded that the cost-effectiveness of universal childhood PCV7 vaccination is uncertain due to the uncertainties arising out of herd immunity effects and serotype replacement. 4 These indirect effects were quantified based on the then most recent (unpublished) observations from the US. 5 6 It argued, however, that the uncertainty, in terms of cost-effectiveness, would be lower using a 2+1 schedule, than the 3+1 schedule, which was recommended at the time by the Superior Health Council. It showed that the incremental cost-effectiveness of using the 3+1 versus the 2+1 schedule was likely unfavourable. In June 2006, the Interministerial conference decided to introduce universal childhood PCV7 vaccination in Belgium, using a 2+1 schedule. This decision led to the inclusion of this option in the regional vaccination programmes from January 2007, with a catch up vaccination for children up to 2 years of age. After PCV7 introduction in Belgium, the overall incidence of invasive pneumococcal disease (IPD) showed only a moderate decrease in young children, in spite of a high vaccine coverage from 2007 onwards. Data from post-licensure IPD surveillance showed a high and rapid impact on vaccine types in all paediatric age groups, but also a concomitant rise in the serotypes that are not contained in the vaccine. 7 The extent to which this rise is due to serotype replacement (i.e. replacement of vaccine serotypes by non-vaccine serotypes) is unclear, and other factors such as secular trends and antibiotic use likely play a role as well. The effectiveness (and cost-effectiveness) of PCV7 in Belgium is therefore questionable.

24 8 Childhood pneumococcal vaccination KCE reports TWO NEW CONJUGATE PNEUMOCOCCAL VACCINES Two new pneumococcal conjugate vaccines (PCV) received EC authorization in These vaccines cover the 7 serotypes included in the PCV7 vaccine, as well as additional serotypes that are now responsible for a high proportion of invasive disease: 1. Synflorix or PCV10 (GSK) is a 10-valent vaccine that also contains containing antigens from the same 7 serotypes than PCV-7, together with capsular polysaccharide antigens from serotypes 1, 5 and 7F. They are conjugated for 8 of them to a surface protein D from H. influenzae and for 2 of them to modified Diphteria toxin and Tetanus toxoid respectively. PCV10 received marketing authorization from the EC in March 2009, and is approved for active immunisation against invasive disease and acute otitis media (AOM) caused by Streptococcus pneumoniae in children from 6 weeks up to 2 years of age. The manufacturer claims a high protective effect on AOM, which has been estimated at 34% in a clinical trial using a 11-valent precursor vaccine. 8 The study suggests that PCV10 provides protection not only against AOM due to pneumococcal serotypes but also against AOM due to non typable Haemophilus influenzae AOM, as the carrier protein is H. influenzae-derived. However, the EMA has only approved PCV10 for AOM due to pneumococcal serotypes. The indication for pneumonia has not been approved by the EMA, but a large clinical trial on pneumonia and AOM is conducted in Latin America and may result in new indications. 2. Prevenar 13 or PCV13 (Pfizer) is a 13-valent (PCV13), which contains capsular polysaccharide antigens from serotypes 1, 3, 5, 6A, 7F and 19A in addition to the 7-valent serotypes, all conjugated to the modified Diphteria toxin. 9 PCV13 received marketing authorization from the EC in December 2009, and is approved for active immunization of children aged 6 weeks to 5 years for the prevention of invasive pneumococcal disease, as well as pneumonia and otitis media. The clinical advantage of each vaccine is difficult to establish. PCV10 may provide a higher protection against AOM, while PCV13 offers a wider coverage of serotypes causing invasive disease (the 6 serotypes additional to PCV7 caused 65% of invasive disease in children <5 years in 2008 compared to 38% for PCV10. 7 In addition, the incidence of serotypes not covered by any of these vaccines also rose. The choice between these vaccines must also be made in a context of uncertainty regarding future serotype replacement, and taking into account vaccine prices.

25 KCE Reports 155 Childhood pneumococcal vaccination 9 2 OBJECTIVES AND RESEARCH QUESTIONS In view of the availability of these new vaccines, the present report aims to estimate the incremental cost-effectiveness of replacing PCV7 by either PCV10 or PCV13 in Belgium. We aim to take into account the indirect effect (herd immunity) and serotype replacement effects. We report results as incremental costs, incremental effects and incremental cost-effectiveness ratios, and report these for numerous scenarios, distinguishing between the various manifestations of clinical disease (this include many explicit comparisons of each of the above type of results with and without the impact on relatively mild disease caused by otitis media (i.e. apart from many sub-presentations of the results, results are also broadly grouped as IPD and pneumonia alone versus IPD, pneumonia and AOM together). This report is organised as follows. Section 3 presents an overview of the search methods for background information from the international literature. In section 4, we present the efficacy and immunogenicity (based on data from trials) and effectiveness (based on post-pcv7 observational studies) of the various pneumococcal conjugate vaccines. In section 5, we describe the recent pneumococcal disease burden (baseline). We also review the international literature on economic evaluations and mathematical models for PCV vaccination (sections 6 and 7). Section 8 contains original research into the effectiveness and cost-effectiveness of various options of use of PCV10 and PCV13 in the Belgian childhood vaccination programme, using a tailor made simulation model.

26 10 Childhood pneumococcal vaccination KCE reports LITERATURE SEARCHES AND METHODS This report builds on the previous KCE report published in In order to update the knowledge base for the current pneumococcal conjugate vaccines, a literature search was undertaken using the broad combined search string pneumococc* AND conjugat* AND (vaccin* OR immun*) in abstract, title or keyword fields of three databases Scopus, ISI Web of Science (SCI and SSCI) and Medline(Pubmed) to retrieve 2226 items of potential interest, which were archived since 1 st January 2006 up to 1 st March 2011 (Figure 1). These items were divided using more focused search criteria to obtain full articles describing vaccine efficacy measures (using the combined search string efficacy OR opsonophagocytic OR immunogenicity ), post-licensure effectiveness (using the single search term effectiveness ) and economic evaluations (see further below for specific search strings). By manual inspection of title and abstract, the publications thus identified were still further refined to distinguish specific observational studies on invasive pneumococcal disease and otitis media (trials were identified before manual inspection of abstracts by using trial as an additional search term). As shown in Figure 2, for the studies on efficacy we distinguish different research lines pertinent to the different vaccine formulations currently licensed.

27 KCE Reports 155 Childhood pneumococcal vaccination 11 Figure 1. Literature search results for publications of interest in Medline(Pubmed), Web of Science and Scopus (1 st January st March 2011) Scopus 1452 Web of Science SCI & SSCI 1793 Pubmed 1622 Exclude duplicates items Contain efficacy, immunogenicity or opsonophagocytic 628 Contain effectiveness 146 Contain economic evaluation search terms (see text) 135 Not part of categories below 418 Original articles 97 Not full articles, or stated reviews 49 Stated reviews and opinion articles 103 Contain trial 99 Mathematical models (humans) 8 Not part of categories below 75 Case control and cohort studies 18 Mathematical models (humans) 4 Not part of categories below 92 Economic evaluations of PCVs (excluding meeting abstracts) 38 Reviews on economic evaluations of PCVs 5

28 12 Childhood pneumococcal vaccination KCE reports 155 Figure 2. Literature search results by vaccine formulation for publications potentially reporting on trials, retrieved through Medline(Pubmed), Web of Science and Scopus (1st January st March 2011) Potential trials 117 Irrelevant vaccines 13 Burden of disease 3 Adult target groups 15 Reviews and methods 14 Studies of interest 72 PCV 7 PCV9 PCV10 PCV11 PCV Trials 17 Observational studies 21

29 KCE Reports 155 Childhood pneumococcal vaccination 13 4 IMMUNOGENICITY, EFFICACY AND EFFECTIVENESS OF CURRENT PNEUMOCOCCAL CONJUGATE VACCINES 4.1 SEVEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV7) PCV7 ( Prevenar ) uses a common carrier protein, cross-reactive material 197 (CRM 197 ), a nontoxic variant of diphtheria toxin for serotypes 4, 6B, 9V, 14, 18C, 19F and 23F. The characteristics of this vaccine as documented up to 2006 are summarized in depth in Oosterhuis-Kafeja et al. 10 Key studies that were used to inform model-based calculations in the previous KCE report on PCV7 4 were publications of (1) the pivotal randomized controlled trial in Northern California ; of (2) a randomized trial in The UK comparing the full 3+1 versus a reduced 2+1 schedule 14 and of (3) the direct and indirect effects of the PCV7 vaccination programme in the USA As shown in Figure 2, half of the studies on pneumococcal conjugate vaccination in children published since 2006, were on PCV7, with 34 out of 72 studies focusing on the other pneumococcal conjugate vaccine formulations Clinical trials efficacy of PCV7 We note a number of key observations from the new PCV7 trials we identified : The efficacy and safety of PCV7 is unaffected when administered concomitantly with a variety of other vaccines, such as meningococcal C conjugate vaccines, measles mumps rubella (MMR) vaccine or Diptheria Tetanus Pertussis (DTP) vaccine combinations. PCV7 reduces nasopharyngeal carriage of vaccine serotypes (VT), and for some serotypes there seems to be less VT carriage occurring with more doses of PCV7 administered. 17 As previously documented 14, a 2+1 schedule of PCV7 (with the two priming doses at 2 and 4 months of age, as in Belgium) resulted in relatively lower immunogenicity responses to serotypes 6B and 23F, compared to a 3+1 schedule. 19 After priming, immunogenicity tended to be less for 6B (32% to 83% of vaccinees 0.35ug/ml) than for 23F (53% to 88.5% of vaccinees > 0.35ug/ml), a tendency which was not observed with 3 dose priming (83-97% and 85-98%, respectively). Note also that with the 3+1 schedule 6B and 23F tended to give slightly lower immunogenicity after priming compared to the other serotypes. After boosting, the antibody concentrations were more generally comparable for all 7 serotypes (but again lower tendency for 6B and 23F in head-to-head comparative trials of 2+1 versus 3+1). In American Indians, vaccine-type pneumococcal carriage is lower among adults and unvaccinated children under 5 years, if they live with a PCV7 vaccinated person (but no such effect is observed for children aged 5-17 years). This is one of the scarce observations of herd immunity (for any infectious disease) documented in a clinical trial. It provides an empirical observation of herd immunity effects on carriage. 22 PCV7 was found to reduce respiratory tract infections compared to placebo, also when given in a 2+1 schedule in a non-randomized open trial. 28 This impact is however time dependent and was in one study 23 observed to be greater for concomitant influenza and pneumococcal conjugate vaccination versus influenza vaccination alone, but only different from placebo during the influenza season.

30 14 Childhood pneumococcal vaccination KCE reports Post-licensure studies effectiveness of PCV7 A comprehensive review of publications which reported on any sort of observable impact PCV7 may have had in any sort of geographical location, is beyond the scope of this HTA report. As outlined above we took a pragmatic approach in identifying first only such studies if they were designed as a case-control study (including indirect cohort or Broome method 33 ) or a cohort study. In addition to this, we also identified 20 publications explicitly stating that they report on the effectiveness of PCV7 (see search terms above) in observational studies Clearly, many other published studies on effectiveness were not retrieved in this way (though many of these are also discussed below), because there is confusion around the term effectiveness (and efficacy and impact) in the literature. We present here the types of vaccine effects (direct, indirect and overall, see Figure 3) and the key studies under each vaccine effect. It is important to note that the indirect and overall effects are defined within the context of a particular intervention program, thus depending on the level of uptake and vaccine allocation within the population. 54 Figure 3. Halloran diagramme on vaccine effectiveness 54

31 KCE Reports 155 Childhood pneumococcal vaccination Direct effectiveness Direct effectiveness measures the direct protective effect of the vaccine in a person who is vaccinated. It compares risks or rates in vaccinated persons and in unvaccinated persons in a population where a fraction is vaccinated. 54 Invasive pneumococcal diseases In the previous KCE report on PCV7, we were fortunate to use pre-publication information on direct effectiveness against IPD from the matched case-control study by Whitney et al. 15 It is noteworthy that Whitney et al 15 found no significant cross protective effect of PCV7 on 19A, but did find a significant effect on 6A. Note though that serotypes 6C and 6D still had to be identified at the time the study was made. A similar, though much smaller, study has recently been published for Canada. 55 Table 1 shows that the results in both studies are in agreement, but that for some specific estimates of direct effectiveness the Canadian study lacked sample size for statistical meaningful interpretation (i.e. regarding a single dose under age 7 months, and specific effectiveness against IPD caused by serotypes 4 and 9V). Barricarte et al 35 reported similar global findings for Navarra (Spain) but with larger confidence intervals due to smaller numbers (overall effectiveness against IPD caused by vaccine serotypes in children under 5 years of age was 88% (95% CI 9-98%)). In Germany, Ruckinger et al 49, used the Broome (case control) method to estimate the direct effectiveness against IPD of one, two and three doses in the first 7 months of life to be 78.1% ( ), 89.8% ( ) and 94.6% ( ), respectively. A US study using the Broome method in Massachusetts state found lower effectiveness values than in Whitney et al 15, at the same period, but had wider confidence intervals: adjusted effectiveness was 90.5% ( ) for the full 3+1 schedule and 76.6% ( ) for 3 doses <7 months of age. 56 Table 1. Comparison of case-control studies in Quebec and the US: % effectiveness of PCV7 against invasive pneumococcal disease (95% CI) Schedule/serotype Quebec; 2-59 months US; 3-59 months Deceuninck et al, Whitney et al, dose 7 months 32% ( ) 73% (43-87) 2 doses 7 months 99% (90-100) 96% (88-99) 3 doses 7 months 90% (24-100) 95% (88-98) 2+1 schedule 100% (82-100) 98% (75-100) 3+1 schedule NA 100% (94-100) Vaccine types serotype 4 72% ( ) 93% (65-99) serotype 6B 90% (49-98) 94% (77-98) serotype 9V 78% ( ) 100% (88-100) serotype 14 98% (84-100) 94% (81-98) serotype 18C 92% (45-99) 97% (85-99) serotype 19F 93% (61-99) 87% (65-95) serotype 23F 82% (-10-98) 98% (80-100) Vaccine related types of special interest serotype 6A 91% ( ) 76% (39-90) serotype 19A 42% (-76-79) 26% (-45-62)

32 16 Childhood pneumococcal vaccination KCE reports 155 Non-invasive pneumococcal diseases Very few studies tried to assess the direct effect of PCV7 on pneumonia and otitis media. O Grady et al 44 found no convincing evidence that PCV7 reduced the incidence of radiologically confirmed pneumonia among subsequent cohorts of indigenous infants in the Northern Territory in Australia, although there was a non-significant trend towards an effect after receipt of the third dose at 6 months of age. 44 Pelton et al 45 compared the difference in effectiveness between a 2+1 and a 3+1 schedule in a matched cohort study, and found a significant higher protection against inpatient treatment for lower respiratory tract infections after the primary series for the 3+1 schedule. There was no significant difference for outpatients, and the difference for inpatients disappeared after the booster dose Indirect effectiveness Indirect effectiveness refers to the population-level effect of widespread vaccination on people not receiving the vaccine. 54 Studies from the US continued to show significant indirect effect among the unvaccinated groups, particularly in the elderly >65 years. Pilishvili documented substantial declines in vaccine types in children too young to be immunized and in adults in 2007 compared to pre-pcv7. 57 Though non-vaccine types also increased in all adults, in particular serotype 19A, the net effect on overall IPD was positive in all age groups >5 years, with reduction in IPD incidence ranging from -18% to -43%. In Europe, indirect effect was also observed but did not result in an overall decline everywhere, because the indirect effect on vaccine type disease was systematically accompanied by a rise of non-vaccine type incidence While the net effect on overall IPD in adults was positive in Norway in 2008 (-15% to -51% depending on age group), 60 it was negative in French adults where IPD incidence significantly increased in all age groups in 2008 compared to pre-vaccine. 58 In the Netherlands and in Navarra (Spain), no substantial changes were measured in older non-immunized children and adults Overall effectiveness Overall effectiveness refers to the overall effect of a vaccination programme on a population, including direct and indirect effects. In most studies, it is measured by comparing the rates in a vaccinated population to those from the same population preintervention (pre and post design). 54 A large number of pre and post studies has been published in many countries around the world, detailing the changing incidence and distribution of the types contributing to disease under the pressure of PCV7 vaccination. These studies confirm that PCV7 is very effective in reducing disease caused by the seven vaccine types, but nearly all studies signal that increasing disease trends caused by non-vaccine types replace vaccine type disease. Extensive international review studies of PCV7 vaccination programmes and their impact is provided by McIntosh & Reinert 61 and Rozenbaum et al. 62 Although they present the only comprehensive overview of such impact studies to date, it is noteworthy that McIntosh & Reinert 61 wrote their overview while being employees of Pfizer, the developer and producer of PCV7, PCV9 and PCV13. Rozenbaum et al received funding from both Pfizer and GSK (the developer and producer of PCV10 and PCV11) to write their review. Invasive pneumococcal diseases In most countries that have used PCV7 as part of a universal childhood programme, the net effect of vaccination on invasive pneumococcal disease remained positive. That is, the reduction of disease caused by the seven vaccine types and other vaccine related types (i.e. often defined as non-vaccine types of vaccine serogroups, that is 6A, 6C, 6D, 9A, 9L, 9N, 18A, 18B, 18F, 19A, 19B, 19C, 23A, and 23B) was larger than the associated increase in disease caused by non-vaccine types The net reduction in IPD is manifested the most in children younger than 2 years of age. We highlight here some specific countries experiences

33 KCE Reports 155 Childhood pneumococcal vaccination 17 THE UNITED STATES In the United States, the use of PCV7 is associated with a large decline in IPD in general between 1998 and 2008 (76% decline in children <5 years; 45% decline over all age groups). More specifically IPD caused by vaccine serotypes (-99.5%) is virtually eliminated in children <5 years of age. 70 Significant indirect reductions (through herd immunity) in overall IPD in all age groups >5 years were observed, particularly in the elderly >65 years. Since 2002 a gradual increase (+29%) in the incidence of non vaccine type IPD has been observed, mainly due to 19A (+253%). Vaccine-related serotype 19A has become the most common cause of IPD in the US, responsible for almost half of IPD episodes in children <5 years of age. The main identified non-vaccine serotypes in IPD include 7F, 22F, 33F and A cause for concern is the emergence of multidrug resistant strains of non-vaccine serotypes 19A. Another concern is the emergence of serotypes 1 and 3, which are often found in pneumonia cases with empyema. CANADA In Canada, the experience with PCV7 is very similar to the US, with near elimination of vaccine types in IPD, a significant decline in overall IPD despite the gradual emergence of types 19A, 3, 22F, 7F, 5 and 15C (in descending order of importance). 71 Of special interest is the Canadian province of Quebec, which was the first region to intentionally introduce universal PCV7 vaccination according to a 2+1 schedule (in December 2004, whereas previously in the US unintentionally a 2+1 schedule was used due to intermittent vaccine shortages between 2000 and 2004). The impact of the vaccination programme there was lower than in the US experience (50% reduction in IPD in children <5 years of age in compared to pre-universal vaccination years vs. a 76% reduction in the US). Serotype replacement occurred mainly through serotype 19A (around half of cases). 55 EUROPE In Europe, PCV7 has been introduced in the universal infant schedule of more than 15 countries usually at vaccine uptake of about 90%, with most introductions taking place in There has been a mix of schedules (see Table 2 for examples and McIntosh et al 61 for a complete overview). Due to the variety of IPD surveillance systems, comparisons should be made with caution. Table 2. Basic characteristics of universal PCV7 programmes in a selection of European countries and Quebec Country Start Schedule Incidence of IPD caused (months) by non vaccine types Quebec Dec , Relatively stable England & Wales Sept , Large increase Netherlands June , 3, Moderate increase France June 2006 a 2, 3, , (after 2008) Large increase Norway July , Stable till 2008 Mild increase in 2009 Belgium Jan , Large increase Denmark Jan , Mild increase (7F) Germany Jan , 3, Relatively stable till Mild increase in a Recommended and reimbursed in 2003 for a large proportion of French infants through definition of their medical and living conditions as presenting higher risk for pneumococcal disease. IPD: invasive pneumococcal disease.

34 18 Childhood pneumococcal vaccination KCE reports 155 While the impact of PCV7 on vaccine type IPD is similar (a decline of around 90 to 95%), the impact on the overall incidence of IPD is clearly lower in Europe (including Belgium) than in the US, irrespective of the schedule that was used. A striking difference is the rapid increase in IPD due to non-vaccine serotypes in Europe, which has a negative influence on PCV7 impact. As a result, most European countries report a lower overall effect on all IPD compared to the US experience. In France, the net decline of IPD (assuming constant incidence) in children under 2 years of age was estimated at 21% in 2006 (95% CI: 10 31%) and 32.5% in , in The Netherlands at 44% (95% CI: 7 66%), in Norway at 52% (95% CI: 31-66%), and in Denmark at 57% (95% CI: 38 71%). In most European countries (including Belgium), there is a rising IPD incidence due to non vaccine types 1, 7F and 19A, but a multitude of other serotypes are also gaining importance in many of these countries. There is emerging dominance of types 1, 5, 7F and 19A in Spain; 1, 7F and 19A in France; 19A in Portugal; 1, 3, 6A, 7F and 19A in Italy; 1, 7F, 19A, 33F and 23F in England and Wales; serotype 7F and other serotypes not present in PCV13 in Denmark; and 7F and 19A in Greece In Norway and Germany, more recent rises in non-vaccine types also involved 7F and 19A Vaccine serotypes are still prevalent in residual IPD in some countries (e.g., in Germany, serotypes 6B, 14 and 23F prevail, along with non vaccine types 1, 3, 6A and 7F). 61 The non-vaccine types that have replaced vaccine types in Europe are partly different from the ones in North America. Assessing the impact of these vaccines is difficult due to the fact that pre-vaccination IPD incidence was not stable. For instance, it has been suggested that serotype 1 evolved independently of vaccination in various European countries, and that serotype 19A showed a mild increase before PCV7 introduction in several countries. 67 BELGIUM In Belgium up to 2008 there was a net IPD incidence decline of 37% in children <2 years and of 15% in children <5 years compared to pre-vaccine, after adjustment for underreporting. However, the overall effect varied according to whether data were adjusted for other factors or not: in children <2 years it was 46% if incidences were adjusted for pre-vaccine trends and 23% if incidences were not adjusted at all (Figure 4). 7 In children <5 years of age, non-vaccine serotypes increased >2-fold. Compared to pre-vaccine, non-vaccine types 1, 7F, 19A, 10A, 12F, 24F and 33F significantly raised in Serotypes 1, 7F and 19A were the most prevalent and represented all together 55% of IPD in Importantly, serotypes that are not included in PCV10 and PCV13 also increased significantly (IRR = 3.18; 95% CI ). Belgian data also show that some non-vaccine serotypes, mainly 1 and 19A, started rising before the introduction of PCV7 (Figure 4 and Figure 5). 7

35 KCE Reports 155 Childhood pneumococcal vaccination 19 Figure 4. Incidence of invasive pneumococcal disease by serotype group in children <2 years of age and linear regression fitted on the prevaccination period (dotted lines) Belgium: Incidence of IPD in <2 years, Incidence (cases /100,000) Vaccine uptake (proxy) Overall IR PCV7 NVT overall fitted VT fitted NVT fitted Number PCV7 schedules per 100 children <2 years Figure 5. Incidence of invasive pneumococcal disease due to serotypes 1, 7F, 19A and 33F by age group

36 20 Childhood pneumococcal vaccination KCE reports 155 There are several limitations with interpreting trends in disease incidence and comparing these between countries. Indeed, in addition to the influence of PCV7 vaccination, there are several relevant aspects to consider when interpreting these trends: Increased reporting has been observed in many countries, as a result of enhanced surveillance that followed introduction of PCV7 in a majority of countries. 78 This becomes a bias when studies do not adjust incidences to the under-reporting rates, as shown by comparison of adjusted and non-adjusted incidences in several EU studies. 2. Increased detection of cases through more frequent blood culturing, especially in cases of bacteremia without focus (see for example ), could contribute to an increase in observed IPD incidence. 3. It is well known that secular trends occur in the pneumococcal serotype distribution. This was documented during the pre-pcv7 vaccination period over both short and long time periods in various countries The pneumococcal serotype distribution can change quickly when different strains start circulating as part of a natural process of evolution (e.g., clones of serotype 1 have often spread rapidly, often remaining a local peculiarity, but also regularly emerging from a local community to become established in wider geographic areas) Antibiotic use differences and changes through time may influence the distribution of serotypes causing disease. For example, although in virtually all countries an increase is observed in drug-resistant clones of serotype 19A, this increase seems to have started in the pre-vaccination era in many of these countries. In Norway, which typically uses fewer antibiotics compared to other countries, the incidence of IPD caused by 19A increased significantly post PCV7, but not that of penicillin-nonsusceptible 19A. 85 Furthermore, several observational studies have documented a positive association between PCV7 vaccination or the number of PCV7 doses received and the likelihood of 19A isolation, including in one of the RCTs These facts suggest effects from both vaccination and antibiotic use on the emergence of (at least) non-vaccine serotype 19A. Note that serotype replacement has not only been documented in relation to IPD and pneumonia, but also in relation to otitis media Given the relatively recent insight that there exists a previously unknown serotype 6C which can replace vaccine type 6A, Millar et al 42 retrospectively showed that over 90% of serogroup 6 invasive pneumococcal disease and carriage strains among Navajo and White Mountain Apache communities have become 6C, a doubling compared to the pre-vaccination era. Note that, in the mean time, now also serotype 6D has been identified. 93 Hanage et al also showed that some cases of serotype 6C can be the results of serotype switching from the vaccine serotype 6B, representing thus escape variants emerging after PCV7 introduction. 94 Pneumonia and otitis media Jardine et al 95 documented reductions in hospitalisations for myringotomy with ventilation tube insertion (MVTI) after the introduction of PCV7 in Australia, of 23%, 16%, and 6% in children aged 0-1years, 1-2 years and 2-3 years, respectively. Mackenzie et al 41 found no significant reduction in otitis media episodes when comparing two cohorts of indigenous infants in Australia, although there was a reduction in tympanic membrane perforation. Ansaldi et al 34 reported declines in all-cause and pneumococcal pneumonia and acute otitis media hospitalisations under 2 years of age in Liguria (Italy) based on comparing ICD-coded hospitalisations three years pre-vaccination with three years postvaccination. Surprisingly, they found no decline in hospitalizations for meningitis or sepsis, but this study had methodological limitations and population and schedule are not comparable to our settings. 34

37 KCE Reports 155 Childhood pneumococcal vaccination 21 Grijalva et al estimated otitis media outpatient visit rates in the US declined by 33% (95% CI, 22%-43%) in children aged <5 years. They also found significant decreases in outpatient visit rates for acute respiratory infections in children aged <5 years. 96 Another US study, using time-series analysis, found a 39% (95% CI 22%-52%) reduction in all-cause pneumonia admission rates in children younger than 2 years. 97 Taylor et al found an overall 24% decrease in OM incidence rates based on insurance data from 9 US regions, but OM rates had already declined 19-24% prior to PCV7 introduction. 98 A long term study (2001-June 2009) from France in children presenting with AOM to a network of paediatricians shows clearly that children who have AOM with fever carry significantly fewer pneumococci (-13% over the first five years), significantly fewer PCV7 serotypes of pneumococci (-61% over the first five years), and significantly more non- PCV7 groups of pneumococci (+145% over the first five years). Over the most recent years ( ), both healthy children and children presenting with AOM (with different associated symptoms, including fever) progressively carried significantly more Haemophilus Influenzae (in 2009: 15.2% in healthy children versus 46.8% in children with AOM), and significantly fewer pneumococci (in 2009: 28.8% in healthy children versus 59.5% in children with AOM). Whereas carriage of 19A increased significantly during the pre-vaccination period, in the post-pcv7 period, there was no significant difference in serotype 19A, and serotype 6A/C carriage (personal communication, Cohen, 2011) Pathogenicity of serotypes By relating IPD incidence to carriage of serotypes during 3 winter seasons ( , , ) in Massachusetts (US), Yildirim et al 69 identified 18C, 33F, 7F, 19A, 3 and 22F as serotypes with the highest invasive capacity, and serotypes 6C, 23A, 35F, 11A, 35B, 19F, 15A and 15BC as serotypes with the lowest invasive capacity. Using data from England & Wales ( ), Trotter et al 99 found that serotype 1 was associated with significantly lower odds of meningitis, and serotypes 23F, 3, 6B, 19F, 18C, 6A, 22F, 12F with significantly higher odds of meningitis relative to serotype 14. Based on case:carrier ratios, they also identified - as the serotypes with the greatest capacity to cause invasive disease - serotypes 7F, 18C, 38, 9V and 14 for children aged less than 5 years, and serotypes 8, 4, 9N, 9V and 3 for everyone aged 5 years and older. These results are to be interpreted with caution since Trotter et al relate the results of local carriage studies with national data on IPD. 99 Using a smaller dataset from Finland ( ), Hanage et al 100 identified serotypes 38, 14, 18C, 19A, and 6B as having a greater risk of causing invasive disease (whereas serotype 6A, 35F and 11A had a lower odds for IPD than average). In the pre-pcv7 period, Brueggemann et al studied the invasive potential of serotypes based on datasets from 6 countries. They identified serotypes 1, 5 and 7 as having a significantly greater odds than serotype 14 of giving rise to IPD once carried, and being 60-fold more invasive that serotypes 3, 6A and 15, those with the lowest odds. 101 They also found that the most invasive serotypes were the least commonly carried. 101 Ruckinger also found that serotype 7F was significantly associated with fatal and severe outcomes (OR>4). 102 Based on these studies, serotypes 18C and 7F would be the most widely recognised as the serotypes with high invasive capacity. Similarly Greenberg et al 103 estimated serotypes with the highest disease potential for paediatric community-acquired alveolar pneumonia (CAAP) to be serotypes 1, 5, 22F, 7F, 14, 9V and 19A, by decreasing rank. In Israel in the period , comparing carriage in healthy children with that in children presenting with IPD and AOM, Shouval et al 104 found serotypes (in descending order of importance) 1, 5 and 12F to be significantly associated with IPD, and serotypes 3, 5, 1, 12F, 19A and 19F with AOM.

38 22 Childhood pneumococcal vaccination KCE reports 155 The above findings do not imply that these types are carried the most in patients with the specific symptoms, just that they are more likely to cause the disease state than other types, if carried. Clearly, if vaccination causes serotype replacement such that these more pathogenic types are more often carried, then there would be a disproportionate increase in the associated disease states. As far as we know, these type-specific aspects have not been considered explicitly in economic evaluation to date (see also below). 4.2 TEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV10) PCV10 contains antigens of serotypes 1, 5 and 7F in addition to the serotypes of PCV7. In PCV10, 8 serotypes are conjugated with a Haemophilus influenzae-derived protein D as a carrier. Serotype 18C is conjugated with tetanus anatoxin and 19F with diphtheria anatoxin Clinical trials immunogenicity of PCV10 We identified 11 publications discussing trials of PCV10 since After inspection of full text articles we listed in Table 3 the five amongst these, which were publications on Randomised Controlled Trials (RCTs), reporting immunogenicity measures. Table 3. Clinical trials used to document immunogenicity of PCV10 Number of Author Countries Design Schedule subjects (PCV10) Vesikari 106 Bermal 112 Wysocki 105 Finland, France, Poland Philippines, Poland Germany, Poland, Spain RCT RCT RCT (18) months weeks months (18) months Number of subjects (PCV7) NA NA (18) months Denmark, (12) 175 NA Silferdal 108 Norway, months RCT Slovakia, (12) 176 NA Sweden months Prymula 109 Czech (15) 226 PP vs RCT NA Republic months 233 NPP RCT: randomised controlled trial; NA: not applicable; PP: prophylactic paracetamol; NPP: no prophylactic paracetamol None of these trials evaluated the clinical efficacy of PCV10. Instead, in order to demonstrate vaccine efficacy, producers of PCV10 have focused on markers indicating an immune response was mounted (by serotype), expressed by the ELISA antibody concentrations, and opsonophagocytic activity (OPA). The latter is a lesser known marker, which provides an in vitro measurement of the ability of serum antibodies to eliminate pneumococci (i.e. it is considered to represent a direct correlate of protection against pneumococcal infection) Furthermore the proportion of subjects with OPA titre >8 is believed to correlate well with pneumococcal vaccine effectiveness. For the purpose of the economic evaluation we undertake in this report, the trials in Table 3 provided basically the following information of general interest. PCV10 induced ELISA antibody responses against all pneumococcal serotypes in the vaccine 106, also when co-administered with other vaccines. 112

39 KCE Reports 155 Childhood pneumococcal vaccination 23 Postvaccination antibody geometric mean concentrations tended to be lower with PCV10 than with PCV7 106, but not significantly and there were some differences between schedules and settings. 112 As with PCV7, after the primary course serotypes 6B and 23F induced the lowest response, relative to the other serotypes. Noninferiority of PCV10 versus PCV7 over the 3 doses primary course could be demonstrated for 8 of the 10 serotypes (not for 6B and 23F), in terms of ELISA antibody response. 106 It could also be demonstrated for 6B and 23F based on the proportion of subjects with OPA titre >8 using a three dose primary schedule. 106 It remains somewhat unclear how reliable the OPA cut-off is to predict clinically relevant efficacy since strong post booster responses were observed for all serotypes in the reduced schedule, and experience with PCV7 seems to contradict that the 2 dose priming schedule would be significantly less efficacious with regard to clinical protection In Wysocki et al 105, who used a schedule which is in line with the Belgian schedule (2 and 4 months for the 2 dose primary course) the ELISA immune responses following 2 primary PCV10 doses were higher for 6B and vaccine related type 6A, and lower for 4, 9V and 18C compared to those following 2 doses of PCV7. They were comparable for the other serotypes. In terms of OPA titres, PCV10 induced higher responses on 6B and 19F, and comparable ones on the other PCV7 vaccine or vaccine related serotypes. 105 The higher response on 19F is important for the PCV10 producer to make the case that cross protection would occur for PCV10, but not (as shown in Whitney et al 15 ) for PCV7. Silfverdal et al 108 compared a 2+1 schedule (which differed from the Belgian schedule, in that the first dose was given at 3 instead of 2 months) with a 3+1 schedule. They noticed that the reduced schedule yields lower post-primary and post booster ELISA antibody levels and OPA titres. Prymula et al 109 investigated the effect of prophylactic paracetamol use around the time of vaccination, and found it to significantly reduce the response on all markers of immunity, on all 10 serotypes in PCV10, and those of jointly administered vaccines. Using these and other data from the same study, Prymula et al 118 showed that PCV10 (like PCV7 and PCV9) reduced nasopharyngeal carriage of vaccine type pneumococci by on average 21.7% compared to PCV7. Non-vaccine types were carried relatively more with PCV10, especially pre-booster, while carriage of other pathogens was unaffected (eg, Haemophilus Influenzae type B). Chevallier et al 111 described safety and tolerability focusing on co-administrations as reported in the different trials in Table 3. These results were more recently confirmed by Bermal et al 119 for different co-administrations. It appears that the following combination vaccines can be administered at the same visit with PCV10: diphtheriatetanus-acellular pertussis vaccine (DTPa), hepatitis B vaccine (HBV), inactivated polio vaccine (IPV), Haemophilus influenzae type b vaccine (Hib), diphtheria-tetanus-whole cell pertussis vaccine (DTPw), measles-mumps-rubella vaccine (MMR), varicella vaccine (V), meningococcal serogroup C conjugate vaccine (CRM 197 and TT conjugates), oral polio vaccine (OPV) and oral rotavirus vaccine. Although there was a tendency for more frequent fever episodes with PCV10, this was not significantly different to PCV7. Knuf et al 110 reported on immunogenicity of other vaccines when they are co-administered with PCV10 or PCV7, and also found no statistically significant reductions in immune response. Vesikari et al 107 reported on the safety and immunogenicity of booster doses in Finland (half the recruits from the above trial by Vesikari et al primed by PCV10, half had been primed by PCV7 and were newly recruited for this new study), using either PCV10 or PCV7 as a booster dose, coadministered with measles mumps rubella varicella (MMRV) vaccine or DTPa-HBV- IPV/Hib (Infanrix hexa) hexavalent vaccine. No significant impacts were found.

40 24 Childhood pneumococcal vaccination KCE reports 155 Focusing on serotypes 19A and 19F, Poolman et al 120 used trial data to confirm that vaccination with either PCV7 or PCV10 induces sufficiently high concentrations of antibodies against serotype 19F, but also to emphasize that higher levels of functional antibodies (OPA) against 19F and 19A polysaccharides were induced by PCV10 than by PCV7. They argued that the conjugation method used in PCV10 is such that it provides better cross protection to serotype 19A than PCV7. Serotype 19A has become an important serotype of residual IPD in countries that introduced universal PCV7 vaccination (see below). Since PCV10 is not widely used yet, it has to our knowledge not been demonstrated to which extent it would induce herd immunity, and whether the herd effects would be comparable to those induced by PCV7 for common serotypes (and given the same conjugation method, also by PCV9 and PCV13). Since PCV10 was shown to reduce nasopharyngeal carriage of vaccine type pneumococci (see above in this section), a basic prerequisite for the potential to induce herd immunity is fulfilled Clinical trials efficacy of PCV10 Of major public health interest is the ability of PCV10 to prevent clinical non-invasive disease in children. The only trials assessing the clinical efficacy of pneumococcal conjugate vaccines of higher valency than PCV10, were conducted using one of the PCV10 precursor vaccines (different formulations of PCV11) aiming to assess its efficacy against AOM In addition to the 10 serotypes in PCV10, PCV11 contained also serotype 3 (apart from this addition there were no substantial differences in vaccine design between PCV10 and PCV11). The immunogenicity and reactogenicity of PCV11 has been described in a series of publications predating , which are summarised in the previous KCE report 4, and Oosterhuis-Kafeja et al. 10 Prymula et al 8 trialled PCV11 in 4968 infants, who were randomly assigned to receive either PCV11 or hepatitis A vaccine at the ages of 3, 4, 5, and months (with follow-up until the end of the second year of life). Parents of children participating the trial were asked to consult their paediatrician if their child was sick, had ear pain, or had spontaneous ear discharge. Children with suspected AOM were then referred to ear, nose, and throat specialists, who functioned as study investigators. They confirmed the clinical diagnosis of otitis media by either the visual appearance of the tympanic membrane (ie, redness, bulging, loss of light reflex) or the presence of middle-ear effusion (by simple or pneumatic otoscopy or by microscopy). Additionally, at least two of the following symptoms were required (within 14 days preceding clinical diagnosis): ear pain, ear discharge, hearing loss, fever, lethargy, irritability, anorexia, vomiting, or diarrhoea. 8 Using this approach, the efficacy of PCV11 against clinically diagnosed AOM was estimated at 33.6% (95% CI %), which was in agreement with a case definition based on such episodes with fever of 38.5C or more (33.9%; 95%CI %). As could be expected, similar or higher estimates were obtained for more specific definitions, such as AOM caused by bacteria (42.1%; 95%CI %), by pneumococcus (51.5%; 95%CI %), by Haemophilus influenzae (35.6%; 95%CI %), and by non-typable Haemophilus influenzae (35.3%; 95%CI %). 8 An excellent overview of the evidence regarding the efficacy of different PCV formulations in reducing AOM is provided by De Wals et al. 132 They modelled vaccine efficacy against all-cause AOM episodes, and found that the most influential factors for differences in vaccine efficacy observed in the different PCV trials were bacterial replacement and the Haemophilus influenzae protein D protection against AOM. Indeed, when they corrected for the prevalence of otopathogens in the control groups of the different trials, they estimated the vaccine efficacy of PCV11 over PCV7 at 10.3 to 11.2% (assuming no serotype replacement), which reduced to %, assuming additionally no impact on Haemophilus influenzae. When they assumed replacement would occur of vaccinerelated otopathogens by other pathogens (to the extent as observed in a Finnish PCV7 trial), they estimated the difference in vaccine efficacy against AOM at %. 132

41 KCE Reports 155 Childhood pneumococcal vaccination 25 The protection of PCV10 against AOM due to non-typable Haemophilus influenzae has also been studied in chinchillas. Immunized chinchillas showed antibody levels against the protein D carrier that were shown to prevent non-typable Haemophilus influenzae AOM, However, the magnitude of the impact of the protein D component remains to be documented. 133 In a serotype specific analysis, Prymula et al 8 found no impact of PCV11 on serotype 3- associated AOM. This finding and the relatively lower immunological response measures for serotype 3, led the PCV11 developer to drop serotype 3 from the formulation, and thenceforth develop, produce and market PCV10. Preliminary results of a large PCV10 efficacy trial (COMPAS) in Latin American countries which had no PCV7 programme, showed a reduction of 7.3% (95%CI %) against suspected community-acquired pneumonia (in-patient and outpatient), and 23.4% (95%CI 9-36%) against radiologically confirmed pneumonia according to WHO criteria Due to the observation that PCV10 yields lower response in 2+1 schedules for some serotypes, and given that significant clinical efficacy of PCV11 was shown only in 3+1 schedules, the European Medicine Agency (EMA) currently licenses PCV10 only under a 3+1 schedule. No data on PCV10 effectiveness are available to date. 4.3 THIRTEEN VALENT PNEUMOCOCCAL CONJUGATE VACCINE (PCV13) The currently licensed PCV13 contains antigens of serotypes 1, 3, 5, 6A, 7F and 19A in addition to the serotypes of PCV7. All antigens are conjugated with CRM 197, the same modified diphtheria anatoxin as is used in PCV7. Building on pre-2006 publications , since 2006 a precursor 9-valent vaccine (PCV9) containing antigens of serotypes 1 and 5 (additional to those in PCV7) was tested in trials in The Gambia and South Africa. 145 Furthermore a combination meningococcal C and PCV9 vaccine was trialled in The UK 146 and Iceland Clinical trials immunogenicity of PCV13 Six publications were identified of trials on PCV As for PCV10, none of these trials assessed the clinical efficacy of PCV13. Since PCV13 was developed with the same technique as PCV7, the trials are less diverse in scope than the trials for PCV10 or PCV7. The PCV13 trials basically demonstrated non-inferiority and safety for PCV13 compared to PCV7, for different schedules and co-administrations, though lower values were observed for 6B in a 2+1 schedule. However, the pivotal trials were conducted with a PCV13 formulation that did not contain the excipient polysorbate 80 (P80) in contrast with the marketed formulation. 9 A bridging study was then conducted to compare the immune responses elicited by PCV13 with and without P80. This study showed lower overall immune response with the P80 formulation, especially for serotypes 6B and 23F. Nevertheless, non-inferiority was met for all 13 valences after the booster injection, but was not met for valences 6B and 23F after primary vaccination. a They also showed that that there are sufficient immune responses for the additional serotypes, both in terms of ELISA immune responses and OPA, and that the efficacy and safety of co-administered vaccines is not adversely affected when given at the same visit as PCV13. However, functional immune responses were lower after the primary schedule for serotypes 1, 3 and 5. It is also noteworthy though that there are indications for some kind of an impaired immune memory response to serotype 3, but the clinical significance of this remains uncertain. a The EMA criteria state Non-inferiority to antibody response for each of the serotypes in the registered vaccine is desirable, but not an absolute requirement. Registration of products in which one or more serotypes do not meet non-inferiority criteria would have to be decided on an individual basis. 9

42 26 Childhood pneumococcal vaccination KCE reports 155 Indeed, serotype 3 exhibited the lowest IgG responses of the 6 additional serotypes after the booster dose, but it exceeded the preset acceptance level of 70%, and the functional OPA antibody levels were comparable, with 98% of subjects exhibiting OPA responses >8 (i.e. indicating that protection is likely). In Kieninger et al 153, the proportions of OPA responders (>8) for serotype 3 were high after both the primary series (99.0%) and the booster dose (98.0%). Although OPA GMT was on average higher after the primary series than after the booster dose, this difference was non-significant at 95% confidence. Furthermore, the immune responses to serotype 3 were compared between a primed (PCV13 primary series + PCV13 booster) and an unprimed (PCV7 primary series + PCV13 booster) group of children. 153 The proportions of children achieving adequate ELISA IgG and OPA responses were similarly high in the two study groups, although the IgG GMCs and OPA GMTs were (non-significantly) lower in the primed group. These observations may indicate that a hyporesponsive state to serotype 3 is not induced by PCV13. Table 4. Clinical trials used to document immunogenicity of PCV13 Author Country Design Schedule (months) Number of subjects (PCV13) Number of subjects (PCV7) Esposito 149 Italy RCT 3, Snape 151 UK RCT 2, Bryant 148 US RCT 2, 4, Yeh 152 US RCT 2, 4,6 + 12(15) Kieninger 153 Germany RCT 2,3,4 + 11(12) Gadzinowski 135 (lot 1) vs. 150 Poland RCT 2,3, (lot 2) NA Study Poland RCT 2, 3, P80 vs P80 NA RCT: Randomised controlled trial; NA: not applicable 13+P80: PCV13 formulated with polysorbate P80: PCV13 formulated without polysorbate 80 In combination with the inefficacy found for serotype 3 AOM prevention with PCV11 8 and the similarities between PCV11 and PCV13 in serotype-specific immunogenicity for common serotypes, the immunological observations on serotype 3 indicate that the PCV13-induced immune response is not sufficient to kill serotype 3 strains in mucosa. However, this does not imply automatically that this insufficiency would also apply to systemic infections like IPD. Currently there is no strong evidence in humans to confirm the latter potential implication. On the contrary, a recent challenge study in rhesus macaques showed good indications that PCV13 would provide protection in humans against IPD with serotypes 1, 3 and 5, even at relatively low OPA titers. 154 A remaining concern with using PCV13 may be the relatively lower immune response for serotypes 6B and 23F observed with 2+1 schedules versus 3+1 schedules for both PCV7 and PCV13. Among the RCTs on PCV13 in Table 4, Snape et al 151 is closest related to the current Belgian schedule; however the PCV13 formulation used in this study did not contain polysorbate 80. In the model input parameter session, we propose adjustments for this observation. No data on PCV13 effectiveness are available to date.

43 KCE Reports 155 Childhood pneumococcal vaccination 27 5 DISEASE BURDEN IN BELGIUM The disease burden before the universal PCV7 vaccination in Belgium is described in the former KCE report and summarized in Table 5. 4 Note that the disease burden of pneumonia and AOM in Table 5 is not limited to that which is definitely related to pneumococcus. Table 5. Estimated annual disease burden of pneumococcal infections prevaccination (2005) in Belgium, all ages pre-vaccination Health state Number of cases IPD infections 1,403 Meningitis 96 Bacteremia 500 Other 807 Deaths Meningitis 12 other IPD 41 Multiple cause pneumonia 608 Life-years lost Meningitis 347 other IPD 663 Multiple cause pneumonia 6,631 Quality-Adjusted Life-Years lost IPD 1,122 Multiple cause pneumonia 6,869 All-cause AOM 1,182 Source: Beutels et al 4. IPD: invasive pneumococcal disease; AOM: acute otitis media This section describes the evolution of the disease burden with S. pneumoniae under the influence of universal PCV7 use in Belgium. The description is mainly based on recent data obtained from Pedisurv (Belgian Institute of Public Health) and the National Reference Laboratory; from the GP sentinel network INTEGO; from the Carenet database of the Christian Mutualities and from the death certificates of the Flemish Community. PCV7 was made part of the regional routine vaccination programmes in Belgium in January 2007, with 2 priming doses at 2 and 4 months of age, and a booster dose at 12 months. Additionally, catch-up vaccination was included for children up to 2 years of age. PCV7 vaccine uptake was estimated at more than 95%, more than 90% and 81-89% for the first, second and final dose, respectively in Wallonia and Flanders in In the following subsections, we describe the evolution of the disease burden in Belgium after the introduction of PCV7, using as much as possible the most recently available data sources.

44 28 Childhood pneumococcal vaccination KCE reports INVASIVE PNEUMOCOCCAL DISEASE (IPD) In Belgium, a national population-based surveillance is conducted in children <16 year olds to monitor post-vaccination pneumococcal epidemiology. This surveillance is based on two prospective and active systems. On the laboratory side, the national reference laboratory (NRL) receives isolates from a stable number of around 100 hospital laboratories distributed all over the country (62% of all laboratories in 2008). It performs serotyping and antibiotic sensitivity testing. On the clinical side, clinical data and vaccination status are collected by a network of paediatricians (Pedisurv) coordinated by the Scientific Institute of Public Health from October 2005 onwards. Cases are matched and these two systems covered ~86% of all confirmed IPD cases in children <5 years in Data are adjusted to under-reporting over time by capturerecapture method. Most of the generated data focused on the evolution of the incidence and serotype distribution of IPD in children. This surveillance showed that the serotype distribution of IPD changed over the years, with PCV7 serotypes disappearing and other serotypes taking their place (i.e. serotype replacement ), thus substantially eroding the impact of PCV7 on the incidence of IPD. 7 There was a net reduction in IPD incidence under 2 years of age, which was less pronounced in >2 years of age. This was also documented in some other European countries (see section 4.1 above), and appears to be in line with the general observation that the use of PCV7 has been more effective in the US 157, and (to a lesser extent) Canada. 55 In particular, serotype 19A has shown a marked and significant rise, representing 28% of all IPD cases <2 years in It rose mainly in the age targeted by PCV7 (<2 years) but presented a mixed picture, increasing before PCV7 use and further rising after vaccination. It should be noted that the prevalence of penicillin and erythromycin nonsusceptible 19A serotypes rose, but 19A susceptible isolates also increased substantially in the same period. This suggests that antibiotic pressure may have played a role but cannot explain alone the 19A rise. This recent 19A trend is thus partly beyond understanding and is likely multi-factorial. The last Belgian data available through Pedisurv and from the national reference laboratory date from The number of cases and the distribution of serotypes for those aged under 16 years is given in Table 6. Table 6. Number and distribution of serotypes causing invasive pneumococcal disease in people under 16 years of age, according to clinical diagnoses (Belgium, 2009) Serotype Menin -gitis Pneumo complic -ated Pneumo non complicated Shock Bacteremia Unknown Other PCV F F B Total PCV F PCV A A Not in vaccine A

45 KCE Reports 155 Childhood pneumococcal vaccination 29 Serotype Menin -gitis Pneumo complic -ated Pneumo non complicated Shock Bacteremia Unknown Other Total 10B A F A B C F F A B B F B F N 1 1 Unknown Total Note: subtotals are additional to previous subtotals (i.e. serotypes for PCV10, are those additional to PCV7, and serotypes for PCV13 are those additional to PCV10). Source: Pedisurv / IPH. Table 6 shows that in 2009, only 10 out of 423 IPD cases aged <16 years with known serotypes were due to serotypes covered by PCV7. An additional 230 cases were due to serotypes 1, 5 and 7F (covered by PCV10 and PCV13), and an additional 86 IPD cases were due to 19A, 3, 6A; all three covered by PCV13 (3 perhaps to a lesser extent); and two of these (6A certainly and perhaps 19A) partially by PCV10 (see also vaccine efficacy sections above). For all IPD, this would bring the theoretical coverage of PCV7, PCV10 and PCV13 in this age group to 2.4%, 56.7% (max 72.6%) and 77.1% (min 74.5%), respectively. For meningitis, the theoretical coverage of PCV10 and PCV13 would be lower and amount to 24.3% and 62.2%, respectively. Yet, as shown in Table 7 a substantial proportion of IPD is caused by serotypes which are not covered by any of the 3 vaccine formulations. Indeed, 14 out of 37 (37.8%) meningitis cases with known serotypes, were caused by non-pcv13 serotypes. Similarly 35 out of 82 (42.7%) bacteremia cases with known serotypes were caused by non- PCV13 serotypes (see Table 6 and Table 7).

46 30 Childhood pneumococcal vaccination KCE reports 155 Table 7. Distribution and invasive capacity of serotypes causing invasive pneumococcal disease in people under 16 years of age, according to clinical diagnoses (Belgium, 2009) Serotypes High Bacteremionia Pneum- invasive Meningitis capacity* % 6B 5.0% - - 9V Y Y C Y F Y 2.5% F 2.5% - - PCV7 PCV10 NON-PCV VACCINE TYPES PCV13 1 Y 5.0% 6.8% 55.7% 5 Y - 8.0% 16.1% 7F Y 7.5% 13.6% 7.4% 3 Not in children 5.0% 3.4% 1.3% 6A 7.5% 2.3% - 19A Y (not more than 19F) 22.5% 19.3% 9.4% 33F Y 10.0% 3.4% 1.3% 22F Y - 8.0% - 38 Y - 4.5% - 24F 5.0% 5.7% 1.3% 12F 5.0% 5.7% 2.0% % 1.1% - 10A - 3.4% - 11A 2.5% 1.1% - other 7.5% 6.8% 4.7% non-pcv % 39.8% 9.4% non-pcv % 69.5% 20.1% non-pcv10 without 19A 51.4% 48.8% 10.7% non-pcv13 with % 46.3% 10.7% * High invasive capacity as identified by Trotter et al 99, Yildirim et al 69 and Brueggemann et al 101, modified by opinions of the expert committee of this report. Source: Perdisurv / IPH In view of the evolution of the serotype distribution observed since the introduction of PCV7 and the potential for rapid rises in IPD caused by non-vaccine pneumococcal serotypes when vaccination exerts ecological pressure on nasopharyngeal colonisation, this observation may add to the uncertainty on the effectiveness of these vaccines in reducing IPD incidence over time. Table 8 summarises the clinical diagnostic information, but distinguishes different age categories to show that the most severe IPD cases (in those aged less than 16 years) occur more often under the age of 1 year and between 1 and 2 years than at higher ages.

47 KCE Reports 155 Childhood pneumococcal vaccination 31 Table 8. Age dependent frequency of clinical diagnoses associated with IPD cases under age 16 years (Belgium, 2009) IPD clinical diagnosis Age <1 [1, 2[ [2, 5[ 5+ Unknown Total Bacteremia Meningitis Other Pneumo complicated Pneumo non complicated Shock Unknown Total Figure 6, Figure 7 and Figure 8 show the IPD incidence per 100,000 population that is minimally covered (broadly and in theory) by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes that occurred in 2007, 2008 and 2009, respectively. It shows that the incremental coverage of PCV10 and PCV13 increases over time. That is, IPD caused by PCV7 serotypes is replaced increasingly by IPD caused by the additional types in PCV10 and PCV13. These figures also show a similar agespecific pattern over all the years. In the group of those aged under 16 years, PCV13 additional coverage compared to PCV10 decreases with age, with the most important gains in coverage expected from PCV13 under the age of 2 years. In the group which shows all ages, the main observation is that the additional coverage offered by PCV10 and PCV13, increases again with age after age 25 years.

48 32 Childhood pneumococcal vaccination KCE reports 155 Figure 6. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2007) Top panel: ages under 16 years; Bottom panel: all ages. Source: Reference Laboratory

49 KCE Reports 155 Childhood pneumococcal vaccination 33 Figure 7. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2008) Top panel: ages under 16 years; Bottom panel: all ages. Source: Reference Laboratory

50 34 Childhood pneumococcal vaccination KCE reports 155 Figure 8. IPD incidence per 100,000 population that is covered by PCV10 and PCV13, incremental to the residual IPD incidence from PCV7 serotypes (2009) Top panel: ages under 16 years; Bottom panel: all ages. Source: Reference Laboratory

51 KCE Reports 155 Childhood pneumococcal vaccination ACUTE OTITIS MEDIA (AOM) AND COMMUNITY ACQUIRED PNEUMONIAE (CAP) PCV7 was expected to be effective against AOM and pneumonia. However, routine surveillance of the Flemish GP sentinel network INTEGO does not suggest evidence of any impact (year of increasing vaccine uptake). The INTEGO surveillance system collects data from about 55 GPs (not fixed per year), working in 47 practices, geographically spread over Flanders. To our knowledge, it is the only validated database that routinely contains incidence data on non-hospital consultations for these ailments. A limiting factor is that it covers only Flanders and not Wallonia and that patients who directly consult a paediatrician are not captured. The data presented here were not collected as part of a study to evaluate PCV7 impact and do not include data from paediatricians. It also covers only outpatient cases, and does not use a standardised clinical case definition nor uses radiological case definitions for pneumonia. Figure 9 for AOM and suspected pneumonia over all ages shows that the incidence per 100,000 population has remained stable after the introduction of PCV7 in Belgium. Figure 9. Evolution of the incidence of GP consultations per 100,000 population due to pneumonia and acute otitis media, in Flanders Source: INTEGO When we observe the age-specific incidence of AOM (Figure 10), we can speculate about the potential reductions PCV10 could produce in this incidence. What appears to be clear by comparing pre-pcv7 data (2004) with post-pcv7 data (averaged over ) is that PCV7 had no impact on all-cause AOM GP consultations in the age group of 0-4 year olds, where since the introduction of PCV7, the incidence had slightly increased (instead of decreased by about 6% (see above)). It is important to keep in mind that this surveillance system was not designed to pick up specific impacts on AOM caused by pneumococcus. Since there are many serotypes for pneumococcus and other pathogens that can cause AOM, the scope for replacement effects seems to be greater than for the other clinical expressions of pneumococcal infections.

52 36 Childhood pneumococcal vaccination KCE reports 155 Figure 10. Evolution of the age-specific incidence of GP consultations per 100,000 population due to acute otitis media and suspected pneumonia, in Flanders Source: INTEGO, HOSPITALISATIONS As explained above, in this section on the disease burden evolution in Belgium, we aimed to study the most recently available evolutions. These were not as up to date in the National database on hospital admissions, as in the alternative Carenet database we consulted in this subsection. We extracted information from the Carenet database for members of the Christian Mutualities (covering about 40% of the Belgian population). All extractions and analyses have been done on anonymised data at the medical direction of the Christian Mutualities, at all times under the supervision of a social insurance physician ( adviserend geneesheer ). At the time of data extraction and cleaning, Carenet contained information on hospitalisations until With about 90% of the hospitals included in this network, Carenet is being considered as representative of the Belgian population Meningitis hospitalisations There is no clearly discernable pattern in the meningitis hospitalisations (all causes) recorded in Carenet, with the possible exception of the youngest age group (under the age of 1 year), in whom the incidence of meningitis hospitalisations steadily declined between 2005 and 2008, but rose again slightly to its 2007 level in 2009.

53 KCE Reports 155 Childhood pneumococcal vaccination 37 Figure 11. Incidence per 100,000 population of hospitalisations with meningitis (all causes) as diagnosis Source: Carenet

54 38 Childhood pneumococcal vaccination KCE reports Bacteremia and septicaemia hospitalisations Figure 12. Incidence per 100,000 population of hospitalisations for patients diagnosed with bacteremia and septicaemia, all causes ( ) Source: Carenet Figure 12 indicates that there appears to be a decline in hospitalisations for bacteremia in those aged between 0-1 years (top panel), 1-2 years and 2-3 years (middle panels). The only other discernable pattern seems to be that in the oldest age group, where the incidence increased (80-90 years, bottom panel).

55 KCE Reports 155 Childhood pneumococcal vaccination Pneumonia hospitalisations 5.4 DEATHS In terms of (all causes) hospitalisations with a diagnosis of pneumonia, there is no discernable trend, neither in overall incidence, nor in the age distribution during the years This is the case for any pneumonia diagnosis, as well as for pneumonia as a primary diagnosis of the hospitalisation Invasive pneumococcal diseases (IPD) deaths Through the Pedisurv network of the IPH we collected information on mortality for patients diagnosed with IPD. There were 1673 cases reported between January 2006 and December 2009 among children younger than 16 years. Information on outcome was available in 1072 cases (64%). Twenty-four children died (case fatality ratio of 2.2%) in this study period. Table 9 shows the evolution of deaths by age group. Children younger than 1 year died mostly following meningitis (73%) while in the older age groups shock, meningitis and complicated pneumonia were diagnosed. Table 9 shows a decrease during the first two years after widespread use of PCV7 in , but there was a rise again in Provisional data indicate that 3 children died in 2010 from pneumococcus. Table 9. Deaths following IPD in children < 16 years old Age Total (2010*) [0, 1[ [1, 5[ [5, 16] Total * Provisional data for Source: Pedisurv, IPH Information on serotype was available for 20 cases. Serotype 7F was observed in 5 cases (25%), whereas serotype 19A and serotype 23F were each found in 3 cases. Table 10 presents the different serotypes by age group. Table 10. Serotype by age group in children < 16 years who died following IPD Serotype Age [0, 1[ [1, 5[ [5, 16] Total A F F F B F V Total Source: Pedisurv, IPH Information on an underlying risk factor for severe IPD was known in 16 of 24 cases and only present in 3 of these cases. Case fatality ratios (CFR) for IPD in children aged <5 years vary significantly in Europe and ranged from 0.7% in Poland to 36.4% in Slovakia, with a mean CFR of 7.4%. 158 In Germany, the case fatality in children younger than 16 years was 4.9% and serotype 7F was associated with a higher risk of severe and fatal outcome than other serotypes. 80 The broad range of CFRs may be related to differences in the patient populations captured by the various surveillance systems, differences in surveillance systems and, to a lesser extent, the differences in healthcare provision between the countries.

56 40 Childhood pneumococcal vaccination KCE reports (Pneumococcal) pneumonia deaths We also obtained information from the communities based on death certificates. As mortality data from Wallonia were incomplete, we focus on Flanders since these death registrations represent uninterrupted time series. However, these data were only available up to 2007 (i.e. at the start of the PCV7 vaccination programme). Figure 13 shows the evolution of deaths with pneumonia and pneumococcal pneumonia as the immediate or underlying cause of death. Figure 13. Deaths with pneumonia and pneumococcal pneumonia as the immediate or underlying cause of death (plotted on left and right hand axis). Source: Death certificates, Flanders Pneumococcal meningitis deaths Table 11 lists the deaths in Flanders from pneumococcal meningitis, indicating that there has been an average of 1 death recorded per year in children 0-3 years of age, between 2002 and Table 11. Deaths in Flanders with immediate or underlying cause recorded as pneumococcal meningitis Age Total [0,3[ Total

57 KCE Reports 155 Childhood pneumococcal vaccination Pneumococcal septicaemia deaths Table 12 shows deaths attributable to pneumococcal septicaemia, based on the death certificate codings. Table 12. Deaths in Flanders with immediate or underlying cause recorded as pneumococcal septicaemia Age Total [0, 6[ [20, 50[ [50, 60[ [60, 70[ [70, 80[ [80, 90[ Total

58 42 Childhood pneumococcal vaccination KCE reports ECONOMIC EVALUATIONS OF PCV10 AND PCV13 A detailed review of economic evaluations predating 2006 is available in Beutels et al. 159 More recent reviews of economic evaluations on PCVs are available, although they each focus on just one particular aspect, such as otitis media 160, herd immunity 161, studies from one country 162, or a small group of countries. 163 By using the combined search string economic OR cost-effectiveness OR cost-benefit OR cost-utility OR cost-effectiveness OR cost-benefit OR cost-utility in our merged database (Figure 1) we identified 38 economic evaluations published since 2006 (excluding meeting abstracts, reviews and cost studies). About 75% of these (or 28 publications) were on PCV In addition to the above 28 cited studies on PCV7, one study estimated the costs and benefits of PCV13 versus PCV7 191, and four others estimated the costs and benefits of PCV10 versus PCV One additional study focused on a novel model to estimate outcomes of PCV10 versus PCV7 while implying vaccination costs would remain the same. 196 Of greater interest than the singular PCV10 and PCV13 studies are four studies, which considered both new vaccine candidates in their analysis. We discuss the English language publications (excluding ) on PCV10 and PCV13 in the next section, in some more detail. 6.1 MODELS STRUCTURE The state of the art of economic evaluation on this subject is to use a static population model to calculate the net indirect effects (of herd immunity and serotype replacement) and either or not combine this with a static cohort model to calculate the direct effects in a single ageing cohort. The combination of a cohort model and a static population model is also the approach that was used by Beutels et al in the 2006 KCE report on pneumococcal vaccination. 4 In structural sensitivity analyses, Beutels et al 4 also modelled the entire population over time (i.e. multiple cohorts), using the 5 years post-pcv7 herd immunity and serotype replacement estimates for the US experience, based on Ray et al. 200 They showed that the difference between these approaches was virtually negligible. None of the published economic evaluations to date have applied a dynamic transmission model to estimate the cost-effectiveness of PCVs. Differences in their results are therefore mainly driven by the input data they used. A notable exception is one group of studies on PCV , which use a steady-state static population model based on a model by De Wals et al 196, assuming that the direct and indirect herd effects of different vaccination options can be estimated over a one year time period from a steady state year in the future, for the total population. It is not clear how the steady state is implicitly defined in these models (i.e. how far in the future the steady state is assumed to be, though implicitly the assumption seems to be that it is reached for PCV7, up to the point where observational data exist). The authors of these studies assert that their approach renders discounting of future costs or effects unnecessary, because they evaluate the cost-effectiveness over a one year period, from the steady state year to the steady state year+1. That is, the future outcomes earned by a series of previously vaccinated cohorts are balanced against the costs of vaccinating a cohort in the present. Yet, the decision maker is investing in the first vaccinated cohort many years before the steady state year. Hence, it is difficult to disentangle time preference effects under this approach. Even if time preference up to the steady state year would not need to be accounted for, by assuming that the steady state year represents an average year in the present, these analyses estimate the number of lifeyears and quality-adjusted life-years (QALYs) foregone due to mortality in the one year time period between the steady state year and the steady state year+1.

59 KCE Reports 155 Childhood pneumococcal vaccination 43 It is incorrect (vis-à-vis current practice of discounting for other health care interventions) to assume that such a future stream of life years should not be discounted back to the steady state year, even if one assumes implicitly that the steady state year is considered as an average year in the present. It is of note that Chaiyakunapruk et al 201 recently made a comparison of PCV7 models, which includes a 2009 version of this steady-state static population model as implemented by GSK ( Supremes ), and some of the more widely used models (combining a birth cohort with a static population model to assess herd immunity). These comparisons show that it is precisely to case-fatality assumptions (and hence life-expectancy) that the steady state model is most sensitive. Hence comparisons between this group of model-based analyses and other approaches should be interpreted with care. 6.2 MAIN ASSUMPTIONS AND RESULTS Since in this report we are interested in PCVs of higher valency, we will not discuss all the PCV7 analyses individually. Many of these were conceptually not substantially different from the PCV7 analysis in the 2006 KCE report. 4 Some of these can be categorised (as in Beutels et al 4 for pre-2006 studies) as irrelevant, since they did not adapt to local serotype distribution and/or did not explore the interplay between the herd immunity effects and serotype replacement and/or did not compare different vaccination schedules. However, some of these publications confirmed and emphasised, as Beutels et al 4 also elaborated in the 2006 KCE report, that the assumed extent of the herd immunity effects and serotype replacement are highly influential, and that the 2+1 schedule dominates the 3+1 schedule, under equally assumed herd immunity effects after the booster dose. A major limitation of these PCV7 analyses is that they were usually based on the US observations regarding herd immunity and serotype replacement. Although it has been amply demonstrated by Ray et al 200 after 5 years and Ray et al 180 after seven years that the net indirect effects were beneficial in the US (with a very cost-effective or even cost saving PCV7 programme as a result), as outlined in section 4.1 above, the net impact of PCV7 on IPD has been different in Europe. Regarding the other analyses on PCV10 and PCV13, the first striking difference is the wide variety in assumed vaccine prices and the lack of scenario analyses with vaccination costs. This was previously shown to be highly influential. Furthermore, none of the studies adjusted the direct vaccine effectiveness by serotype immunogenicity, making the assumption that the additional effectiveness on IPD caused by the additional serotypes would be the same as (sometimes the average) for the 7 serotypes in PCV7 (as shown in Table 13, most assumed this to be 94% for the additional serotypes). Studies attempting to estimate the effects on AOM and pneumonia have assumed that the distribution of IPD serotypes represents well the serotype distribution found in patients with AOM and pneumonia (see also below). Since the effectiveness of PCV10 against AOM remained largely unexplored, head-tohead analyses tended to conclude that PCV13 was more cost-effective than PCV10 (though Chuck et al 197 noted this is reversed, if they assumed an effect on NTHI AOM). The influence of including or excluding cross reactive serotypes is rarely explored. In Talbird et al cross-protection as estimated by Whitney et al 15 for PCV7 was included for 19A and 6A, but the influence of this assumption was not explored. In the 2006 KCE report, Beutels et al 4 used 3 different sets of QALY/DALY estimates (with negligible differences in impact). These sets were also used by the economic evaluations in Table 13 (see below, section model input data). Estimates of quality of life as used by O Brien et al 192, stem from an earlier study by Prosser et al 205 which produced estimates that are very high by comparison to other studies, due to their approach (as previously discussed in Beutels & Viney 206 ).

60 44 Childhood pneumococcal vaccination KCE reports 155 First author, country, year O Brien 192, US, 2009 Chuck 197, Alberta, 2010 Rozenbaum 181, Netherlands, 2010 Rubin 191, US, 2010 Talbird, , Canada, Germany, Mexico, Norway, 2010 Table 13. Main assumptions and results of economic evaluations on PCV10 or PCV13 Vaccination costs per dose schedule PCV7= 45.6 PCV11= schedule PCV7 =PCV10 =PCV13 = schedule PCV7 = 50 PCV10= b PCV13 = schedule PCV7 = 51.6 PCV13 = schedule PCV7=PCV schedule 2+1 schedule Country dependent Model Main effectiveness assumptions Results Markov state transition model Steady-state population model Cohort model Markov state transition model Steady-state population model Effectiveness versus AOM, PCV11 vs. PCV7: months: 28% vs. 6.4% months: 14% vs. 3.2% months: 7% vs. 1.6% IPD and pneumonia effectiveness as for PCV7 in Lieu et al, % against additional serotypes (no cross protection) 3% or 5% a for AOM with NTHI 97.4% against additional serotypes (no cross protection) Net indirect effects of PCV10 = PCV13 = 10% No effect on NTHI AOM Pneumonia and AOM estimates: same as PCV7 (adjusted for additional IPD serotypes) 94% against IPD additional serotypes (no cross protection) 7.3% against suspected pneumonia % against hospitalised pneumonia 4-6.7% against all cause AOM (Fin OM and KCP) 6-10% against complex AOM IPD schedule effectiveness and serotype specific effectiveness (based on Whitney 15 including cross protection (6A, 19A) + 94% against additional PCV10 serotypes). Herd immunity for IPD (PCV10=PCV7): - < 5years: 15.4% - 5years: 29% All-cause pneumonia: - Hospitalised: 25.0% vs. 20.5% (PCV10 vs. PCV7) - Non-hospitalised: 4.3% (PCV10=PCV7) AOM due to VT: 57.6% vs. 57.2% (PCV10 vs. PCV7) AOM due to NTHI: -11% vs % (PCV10 vs. PCV7) Cost-saving with savings from AOM outweighing savings from other pneumococcal disease (includes indirect costs of parental time and QALY losses of parents) PCV13 dominates PCV10 (but reversed if effectiveness against NTHI AOM is included) PCV10 vs. PCV7: 52,947 per QALY gained PCV13 vs. PCV7: 50,042 per QALY gained PCV13 vs. PCV7: cost saving Catch up 16-23m: 2404 per QALY gained Catch up 16-35m: 17,715 per QALY gained Catch up 16-59m: 52,028 per QALY gained PCV10 vs. PCV7: cost-saving

61 KCE Reports 155 Childhood pneumococcal vaccination 45 First author, country, year Sartori 194, Brazil, 2010 Kim 198, The Gambia, 2010 Vaccination costs per dose schedule PCV10= schedule PCV7= PCV10= PCV13= schedule Model Main effectiveness assumptions Results ProVac model (25 cohort model) 204 Markov cohort model All-cause AOM: 6.7% vs. 22.9% (PCV10 vs. PCV7) 94% against vaccine serotypes (no cross protection) for 5 years AOM due to VT: 57% Pneumonia due to VT: 87.5% (All-cause) primary endpoint pneumonia: - PCV7: 26% - PCV10: 35% - PCV13: 41% Pneumococcal meningitis/sepsis: - PCV7: 16% - PCV10: 22% - PCV13: 26% AOM impact excluded Herd immunity, serotype replacement explored in sensitivity analysis PCV10 vs. PCV7 in high risk children: 9,392 per DALY averted All PCVs considered cost-effective (costs per DALY averted < 3xGDP/capita) PCV13 more cost-effective than PCV10, and PCV7 a. Unclear due to conflicting statements b. Average between PCV7 and PCV13 AOM: acute otitis media, VT: vaccine type; NTHI: non-typable Haemophilus influenzae; IPD: invasive pneumococcal disease; CAP: community acquired pneumonia; QALY: quality-adjusted life-year; DALY: disability-adjusted life-year; GDP: gross domestic product

62 46 Childhood pneumococcal vaccination KCE reports MATHEMATICAL MODELS OF PNEUMOCOCCAL TRANSMISSION Our search identified 11 different publications , which presented or discussed mathematical models without being coupled to economic evaluation. None of these models apply to PCV10 or PCV13. Interestingly, Van Effelterre et al 208 focused on the interaction between ecological pressures from PCV7 vaccination and antibiotic use to explain the rising incidence of serotype 19A in the US. This model is designed to generate hypotheses leading to a better understanding of mechanisms. Only two other models represent an application to the transmission dynamics of PCV , with the remainder discussing or comparing different static models , or demonstrating specific transmission dynamic features of pneumococci in theory , like previous pre-2006 models did as well. Melegaro et al 213 and Snedecor et al 218 are the only publications that tried to quantify the impact of the PCV7 at the level of the population. They both developed an age structured compartmental deterministic model. Snedecor et al 218 did not distinguish vaccine serotypes from non-vaccine serotypes. Their approach focused on reproducing the observed herd immunity impact in the US (and exploring the likely impact of different vaccine uptake scenarios), by distinguishing IPD and nasopharyngeal carriage as pneumococcal infection states. Snedecor s et al approach did not allow to account for serotype replacement. In order to fit the observed data on herd immunity (US, ), they departed from the mass action principle used widely in transmission dynamic modelling and defined the force of infection in the oldest age group to be proportional to the squared number of carriers in the youngest age group. The more recent model by Melegaro et al, is able to cope better with some of these methodological challenges. In Melegaro et al 213 serotypes are grouped in either PCV7 vaccine type or PCV7 non-vaccine type groups. They made projections of different PCV7 vaccination strategies and used the following datasets: 1. Longitudinal pneumococcal carriage data from nearly 500 individuals of all ages in England and Wales. These data were used to estimate the force of infection of vaccine serotypes and non-vaccine serotypes. 2. Age specific IPD incidence data by serotype (all ages) from the national surveillance system in England and Wales. In combination with the previous dataset, these data were used to derive case:carrier ratios. 3. The degree and duration of vaccine effectiveness, as well as the level of competition between vaccine and nonvaccine serotypes was derived from US IPD incidence and vaccine uptake data during the pre- and post-vaccination era ( ). The transmission model was fitted to the incidence data (after deriving age-specific mixing patterns from them), by modifying the combination of parameters expressing the age-specific duration of carriage (assumed to be independent of serotype), the degree and duration of vaccine protection against carriage of vaccine and non-vaccine serotypes, as well as the level of competition between these groups of serotypes (and keeping the likely most suitable mixing patterns, derived separately, and vaccine uptake estimates fixed). After the main parameter sets were decided on, projections were made for England and Wales suggesting that vaccine type pneumococci would be eliminated over a 5 to 10 year time period, but exhibiting at the same time high sensitivity to the parameter values that express competition between vaccine and non-vaccine serotypes. As expected, the duration to elimination of vaccine types is shorter with more expansive catch-up programmes (to age 60 months). The predictions related to nonvaccine types remained much more speculative and highly sensitive to the selection of fitted parameter sets for carriage duration, the force of infection of and competition between vaccine types and non-vaccine types. The latter is also the main uncertainty in observational studies, and Melegaro et al duly noted that estimates of replacement effects are affected by secular trends in prevalence and antimicrobial sensitivity of

63 KCE Reports 155 Childhood pneumococcal vaccination 47 serotypes at the time of introduction, differences in surveillance systems, differences in clinical practice. None of these aspects could be taken into account in their study. In addition to the limitation that the approach requires to calibrate the model by changing many different parameters simultaneously (and hence the options for fitting are tremendously large, despite the fact that social mixing is kept constant), they discussed the following main limitations of their projections: 1. They included serotype 6A in the PCV7 vaccine type group (because of the observed cross protection), but this was using data from the time before 6C and 6D had been identified, and this may affect the fittings and projections. 2. They grouped all serotypes in two main groups, while there exists considerable heterogeneity between different serotypes in each group (e.g., transmissibility, duration of carriage, ability to co-colonise, ability to prevent co-colonisation with other serotypes, ability to cause disease). Despite that Melegaro et al 213 have made impressive progress in this area for PCV7 (their main aim is to support projections for policy in developing countries), in the current policy supporting report we chose not to undertake the first dynamic transmission model approach of PCV10 or PCV13 for four main reasons: 1. While we have data for Belgium on carriage of pneumococci in children (from one cross sectional study in day care centres pre-pcv7 introduction 221, and one cohort study in Brussels schools during the period of PCV7 introduction ( ) 222 ), this is currently lacking for the general population in Belgium. 2. The IPD incidence data for Belgium were not available by serotype for adults at the start of our study (only in children). Data on serotype-specific IPD in adults over 50 years of age have become recently available to us (2011). 3. We believe that a number of fundamental aspects of colonisation, carriage and transmission of pneumococci are still poorly understood, such that the role of transmission models would be more to help understand and generate hypotheses, rather than to make projections to support policy making. For the latter, our lack of understanding basic mechanisms to model transmission and colonisation undermines our ability to quantify or even identify uncertainties that are embedded in these types of projections. 4. Last but not least, in order to undertake such a complex modelling study, we would require research capacity that surpasses the time and resources typically available to undertake an individual KCE report. As a research agenda to enable and improve dynamic models for pneumococci, we propose the following: 1. Apply Melegaro et al s approach for calibration to different European countries instead of the US (in Melegaro et al 213, the model is fitted to US data, and hence the implicit assumptions of their model (e.g., regarding antibiotic use) are assumed to be transferable to England and Wales). 2. Develop laboratory techniques that can detect carriage of multiple serotypes from nasal aspiration or swaps. 3. Undertake a representative general population study in Belgium (oversampling children, preferably sampling full households) to obtain pneumococcal carriage information (e.g. by nasopharyngeal swaps or aspiration) at regular intervals over a period of a year. Ideally techniques would be available to detect multiple serotypes being carried simultaneously 4. Enhance IPD surveillance, as already done in children, such that (at least over the same period as above) IPD cases of all ages are serotyped (also including multiple serotypes), and clinical information and antibiotic susceptibility of serotypes is determined 5. Provide long term support to undertake and build capacity in basic research in mathematical modelling of infectious diseases in humans

64 48 Childhood pneumococcal vaccination KCE reports COST-UTILITY ANALYSIS OF PCV10 AND PCV13 IN BELGIUM The general principles of this analysis are that we parameterise the uncertainty we can parameterise using distributions, and that we model distributions and present the results as distributions. We explore the assumptions we make in order to do this. We show all outcomes separately for IPD (distinguishing meningitis, bacteremia and other IPD), otitis media and pneumonia. We show all results for different levels of protection offered by PCV10 against AOM, distinguishing between AOM caused by Non Typable Haemophilus Influenzae (NTHI) and all-cause AOM. We show all results with and without herd immunity, and we explore the impact of serotype replacement. We perform multivariate threshold analysis on price differences between different vaccine formulations. We show the impact on effectiveness and cost-effectiveness of using three different interpretations regarding the correlates of protection for the efficacy of PCV10 and PCV13, and show the impact of including or removing individual serotypes for which protection can be contested. In order to be able to do this, we needed to develop a flexible model that can incorporate probabilistic sensitivity analysis. As explained in the review section on mathematical models, for various reasons we opted to attempt this by using a static model, in which herd immunity and serotype replacement is imposed rather than estimated through the dynamics of transmission of pathogens. We do not include the impact vaccination may have on preventing antibiotic resistance. 8.1 STUDY DESIGN General Data analyses and simulations were performed using MS Excel 4.5 and SAS. The baseline costing perspective is that of the Belgian health care payer, which includes collective payments by the Belgian health care system, as well as co-payments for health care by patients. All cost data are expressed in Euro Our primary measure of relative efficiency is direct costs per Quality-Adjusted Life-Year (QALY), though a wider range of health outcomes is presented in incremental cost-effectiveness analyses. Time preference is accounted for by discounting costs at an annual constant rate of 3%, and effects at 1.5%. These analytical choices are in line with the Belgian guidelines for economic evaluation in health care 223 and an international WHO guide on economic evaluations of vaccines Vaccination options The options for vaccination were selected based on global experience with PCV7, as well as the results from clinical trials with PCV10 and PCV13 (see sections 4.2 and 4.3 above). Given the routine Belgian infant vaccination schedule and the fact that PCV10 is only licensed under a 3+1 schedule, we focus on the following options in our analysis: Option 1: Current situation. PCV7 vaccination using a 2+1 schedule with injections at 2, 4 and 12 months of age. Option 2: PCV10 vaccination using a 3+1 schedule with injections at 2, 3, 4 and 12 months of age. In addition, the 2+1 schedule has been considered, in anticipation of potential changes in the authorized schedule. Option 3: PCV13 vaccination using either a 2+1 or 3+1 schedule with injections at 2, (3), 4 and 12 months of age. Options 2 and 3 are compared to option 1, as well as incrementally to each other.

65 KCE Reports 155 Childhood pneumococcal vaccination Mathematical model structure The economic analysis is by necessity based on a mathematical simulation model. We have opted for an integrated model, which combines two submodels, and can be subjected to multivariate probabilistic sensitivity analysis Static cohort model without herd immunity An age-structured classic Markov model was developed in MS Excel, simulating costs and effects of pneumococcal disease and vaccination in a single closed Belgian cohort followed from birth until 100 years of age. The first 6 years of the model run in monthly cycles. The following 94 years run in annual cycles. The model is flexible in that any 1 to 4 dose schedule can be assumed over the 100 year time span (and the time span can be adjusted), and specific in that in under 6 year olds, the timing of each dose can be focused on any month. This model is static, so like all the other models used for economic evaluation on pneumococcal conjugate vaccines (see section 6 above), it does not generate herd immunity effects, based on built-in transmission dynamics. Figure 14. Basic structure of the static cohort model 1 Not vaccinated, or vaccinated but not (or no longer) protected 2 Modelled in separate states (and with separate transitions from and to other states) for IPD, allcause pneumonia and all-cause otitis media (detail not shown for clarity) Static population model to assess indirect effects for PCVs Based on observations in the USA and Belgium, the indirect effects (of herd immunity and serotype replacement) are simulated (and their consequences in terms of costs and effects) in a static population model. This population model assesses one year of infections over the entire population to simulate herd immunity and serotype replacement at the population level. Herd immunity effects for PCV as estimated in the static population model are looped back into the static cohort model, so that in fact both submodels form one integrated model. This implies also that sensitivity analysis is performed on the integrated model. Since the impacts of herd immunity and serotype replacement are highly speculative the uncertainty around these aspects are not only parameterised in the models, but are also elaborated in scenario analyses (see results section 8.3). As recommended by international guidelines for the economic evaluation of vaccination programmes , the time span is chosen during the analysis such that the median ICERs reach a plateau. In the current report the focus is on a 5 year time span for incurring infections, and a lifelong time horizon for infected persons incurring long term consequences (sequelae including death) from their infections. Alternative calculations using a 10 year time span are also shown in sensitivity analyses.